` (the figure's name) and a `<desc>` (the alt text from the brief).\n- **Size budget.** Under 20 KB with system fonts; under 64 KB with an embedded font subset. No raster images, no hidden layers, no editor metadata. Coordinates rounded to one decimal.\n- **Stored beside the post** as `blog/<slug>/figure.svg` (or `<folder>/figure.svg` for other articles), and linked from the markdown as\n\n ```markdown\n ![<alt text: what is shown and what it shows, 1–3 sentences>](figure.svg)\n\n *<caption: the takeaway in one sentence>*\n ```\n\n One figure per article by default. A second figure must pass Step 1 on its own and is named for what it shows (`figure-scopes.svg`).\n\nA starting skeleton:\n\n```xml\n<svg xmlns=\"http://www.w3.org/2000/svg\" class=\"tf\" viewBox=\"0 0 600 400\" width=\"600\" height=\"400\"\n role=\"img\" aria-labelledby=\"t d\">\n<title id=\"t\">Where the text of one file goes

[{"_1":2,"_3":-5,"_4":-5},"loaderData",{"_5":6,"_7":8},"actionData","errors","root",{"_248":249},"routes/registry-package",{"_9":10,"_11":12,"_13":14,"_15":16,"_17":18,"_19":20},"pkg",{"_26":27,"_28":29,"_30":27,"_31":32,"_33":34,"_22":23,"_35":36,"_37":38,"_39":40,"_41":42,"_24":25,"_43":44},"comments",[],"versions",[21],"selectedFile","tufte-charts/requirements.txt","canonical","https://agentlefs.com/registry/skills/tufte-charts/requirements.txt","historical",false,{"_22":23,"_24":25},"version",1,"publishedAt","2026-10-19T06:25:38.587Z","slug","skills","kind","document","title","summary","","publisher","agentleFS","body","MIT License\n\nCopyright (c) 2026 agentleFS\n\nPermission is hereby granted, free of charge, to any person obtaining a copy\nof this software and associated documentation files (the \"Software\"), to deal\nin the Software without restriction, including without limitation the rights\nto use, copy, modify, merge, publish, distribute, sublicense, and/or sell\ncopies of the Software, and to permit persons to whom the Software is\nfurnished to do so, subject to the following conditions:\n\nThe above copyright notice and this permission notice shall be included in all\ncopies or substantial portions of the Software.\n\nTHE SOFTWARE IS PROVIDED \"AS IS\", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR\nIMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,\nFITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE\nAUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER\nLIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,\nOUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE\nSOFTWARE.\n","fields",{},"files",[49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78,79,80,81,82,83,84,85,86,87,88,89,90,91,92,93,94,95,96,97,98,99,100,101,102,103,104,105,106,107,108,109,110,111,112,113,114,115,116],"approvedBy","ContextHub's Team Org admin","standing",{"_45":46,"_47":46,"_48":-5},"chains",0,"ratings","meanByChain",{"_117":247,"_119":36},{"_117":245,"_119":246},{"_117":244,"_119":36},{"_117":242,"_119":243},{"_117":240,"_119":241},{"_117":238,"_119":239},{"_117":236,"_119":237},{"_117":234,"_119":235},{"_117":232,"_119":233},{"_117":230,"_119":231},{"_117":228,"_119":229},{"_117":227,"_119":36},{"_117":225,"_119":226},{"_117":223,"_119":224},{"_117":221,"_119":222},{"_117":219,"_119":220},{"_117":217,"_119":218},{"_117":215,"_119":216},{"_117":213,"_119":214},{"_117":211,"_119":212},{"_117":209,"_119":210},{"_117":207,"_119":208},{"_117":205,"_119":206},{"_117":203,"_119":204},{"_117":201,"_119":202},{"_117":199,"_119":200},{"_117":197,"_119":198},{"_117":195,"_119":196},{"_117":193,"_119":194},{"_117":191,"_119":192},{"_117":189,"_119":190},{"_117":187,"_119":188},{"_117":185,"_119":186},{"_117":183,"_119":184},{"_117":181,"_119":182},{"_117":179,"_119":180},{"_117":177,"_119":178},{"_117":175,"_119":176},{"_117":173,"_119":174},{"_117":172,"_119":36},{"_117":170,"_119":171},{"_117":168,"_119":169},{"_117":166,"_119":167},{"_117":164,"_119":165},{"_117":162,"_119":163},{"_117":160,"_119":161},{"_117":158,"_119":159},{"_117":156,"_119":157},{"_117":154,"_119":155},{"_117":152,"_119":153},{"_117":16,"_119":151},{"_117":149,"_119":150},{"_117":147,"_119":148},{"_117":145,"_119":146},{"_117":143,"_119":144},{"_117":141,"_119":142},{"_117":139,"_119":140},{"_117":137,"_119":138},{"_117":136,"_119":36},{"_117":134,"_119":135},{"_117":133,"_119":36},{"_117":131,"_119":132},{"_117":129,"_119":130},{"_117":127,"_119":128},{"_117":125,"_119":126},{"_117":123,"_119":124},{"_117":121,"_119":122},{"_117":118,"_119":120},"path","watercolor-data-art/scripts/render.mjs","content","// render.mjs DATA.json OUT.png [--style grid|rothko|squares|matrix] [--seed 20261010] [--size 2400x840]\n// [--palette kind|earth|sea|dusk|ink|\"#a33a2c,#4f7392,...\"]\n// [--ghosts running|global|none] [--cell 26] [--max-scale N]\n// [--background paper|transparent|dark]\n//\n// --background paper (default) is warm paper with grain. transparent leaves the canvas clear and\n// dark paints a flat #111 ground; both skip the grain and lift the pigments so they read on a\n// dark page. All three give the same squares for the same seed, so a pair lines up.\n//\n// --max-scale N caps px per unit at N, sizing by the stack alone (the older rule). Without it the\n// picture is fitted to the canvas, ghosts included. Images made before the auto-fit need it:\n// grid --max-scale 54, rothko --max-scale 50.\n//\n// Paints stacked or matrix data as watercolor in headless Chromium and saves the canvas pixels\n// (not a screenshot) as PNG. Same data + options gives the same picture on the same Chromium\n// build; another GPU or Skia build can differ in antialiasing.\n//\n// Playwright is looked up in this order:\n// 1. PLAYWRIGHT_MODULE: a package name or a path to an installed package folder\n// 2. \"playwright\", then \"@playwright/test\", resolved from the current working directory\n// (so `npm i playwright` in the folder you run from works)\n// 3. the same names in each NODE_PATH folder, then next to this script\n// Chromium: CHROMIUM_PATH if set (any installed Chromium or Chrome); else Playwright's own build;\n// else any Chromium Playwright downloaded earlier (PLAYWRIGHT_BROWSERS_PATH, ~/.cache/ms-playwright).\n// A plain `npm i playwright` installs the newest version, which may want a Chromium build you do\n// not have: pin the version that matches your browsers (e.g. npm i playwright@1.56.1) or set\n// CHROMIUM_PATH. Pixel-identical output needs the same Chromium build.\nimport { createRequire } from \"node:module\";\nimport { pathToFileURL, fileURLToPath } from \"node:url\";\nimport fs from \"node:fs\";\nimport os from \"node:os\";\nimport path from \"node:path\";\n\nfunction loadPlaywright() {\n const here = path.dirname(fileURLToPath(import.meta.url));\n const bases = [process.cwd(), ...(process.env.NODE_PATH || \"\").split(path.delimiter).filter(Boolean), here];\n const names = process.env.PLAYWRIGHT_MODULE ? [process.env.PLAYWRIGHT_MODULE] : [\"playwright\", \"@playwright/test\"];\n const tried = [];\n for (const base of bases) {\n const req = createRequire(path.join(path.resolve(base), \"noop.js\"));\n for (const n of names) {\n try { return req(n); } catch { tried.push(`${n} from ${base}`); }\n }\n }\n console.error(\"Cannot find Playwright. Tried:\\n \" + tried.join(\"\\n \") +\n \"\\nFix: run `npm i playwright && npx playwright install chromium` in the folder you run from,\" +\n \"\\nor set PLAYWRIGHT_MODULE=/path/to/node_modules/playwright (and CHROMIUM_PATH=/path/to/chrome if needed).\");\n process.exit(1);\n}\n\n// Chromium builds Playwright has downloaded before, newest first: used when the installed\n// Playwright wants a build that is not there (e.g. a newer `npm i playwright` than the browsers).\nfunction installedChromiums() {\n const roots = [process.env.PLAYWRIGHT_BROWSERS_PATH, path.join(os.homedir(), \".cache\", \"ms-playwright\"),\n path.join(os.homedir(), \"Library\", \"Caches\", \"ms-playwright\")].filter(Boolean);\n const rel = [\"chrome-linux/chrome\", \"chrome-linux64/chrome\", \"chrome-mac/Chromium.app/Contents/MacOS/Chromium\",\n \"chrome-mac-arm64/Chromium.app/Contents/MacOS/Chromium\", \"chrome-win/chrome.exe\"];\n const found = [];\n for (const r of roots) {\n let dirs = [];\n try { dirs = fs.readdirSync(r).filter(d => /^chromium-\\d+$/.test(d)); } catch { continue; }\n dirs.sort((a, b) => Number(b.split(\"-\")[1]) - Number(a.split(\"-\")[1]));\n for (const d of dirs) for (const x of rel) { const p = path.join(r, d, x); if (fs.existsSync(p)) found.push(p); }\n }\n return found;\n}\n\nasync function launch(pw) {\n if (process.env.CHROMIUM_PATH) return pw.chromium.launch({ executablePath: process.env.CHROMIUM_PATH });\n try { return await pw.chromium.launch(); } catch (e) {\n for (const p of installedChromiums()) {\n try { const b = await pw.chromium.launch({ executablePath: p }); console.error(`note: Playwright's own Chromium is missing; using ${p}`); return b; } catch {}\n }\n console.error(\"Playwright found, but Chromium did not start:\\n\" + String(e.message).split(\"\\n\")[0] +\n \"\\nFix: npx playwright install chromium (with the same Playwright version), or set CHROMIUM_PATH\" +\n \"\\nto any installed Chromium or Chrome binary.\");\n process.exit(1);\n }\n}\n\nconst args = process.argv.slice(2);\nconst opt = (name, dflt) => { const i = args.indexOf(name); return i >= 0 ? args.splice(i, 2)[1] : dflt; };\nconst style = opt(\"--style\", undefined);\nconst seed = Number(opt(\"--seed\", \"20261010\"));\nconst size = opt(\"--size\", \"2400x840\").split(\"x\").map(Number);\nconst palArg = opt(\"--palette\", undefined);\nconst ghosts = opt(\"--ghosts\", \"running\");\nconst cell = opt(\"--cell\", undefined);\nconst maxScale = opt(\"--max-scale\", undefined);\nconst background = opt(\"--background\", \"paper\");\nconst [dataPath, out] = args;\nconst STYLES = [\"grid\", \"rothko\", \"squares\", \"matrix\"];\nif (!dataPath || !out || (style && !STYLES.includes(style)) || (maxScale !== undefined && !(Number(maxScale) > 0)) || ![\"running\", \"global\", \"none\"].includes(ghosts) || ![\"paper\", \"transparent\", \"dark\"].includes(background) || size.some(n => !(n > 0))) {\n console.error(\"usage: render.mjs DATA.json OUT.png [--style grid|rothko|squares|matrix] [--seed N] [--size WxH]\\n\" +\n \" [--palette kind|earth|sea|dusk|ink|#hex,#hex,...] [--ghosts running|global|none] [--cell PX] [--max-scale N]\\n\" +\n \" [--background paper|transparent|dark]\");\n process.exit(2);\n}\nlet data;\ntry { data = JSON.parse(fs.readFileSync(dataPath, \"utf8\")); }\ncatch (e) { console.error(`cannot read ${dataPath}: ${e.message}`); process.exit(2); }\n\n// check the input before starting a browser: every value a number or null, no ragged rows\nfunction validate(d) {\n const bad = [];\n const isVal = x => x === null || (typeof x === \"number\" && Number.isFinite(x));\n if (Array.isArray(d.matrix)) {\n if (!d.matrix.length || !d.matrix.every(Array.isArray)) return [\"matrix must be a list of rows (lists)\"];\n const w = d.matrix[0].length;\n d.matrix.forEach((r, i) => { if (r.length !== w) bad.push(`matrix row ${i} has ${r.length} cells, row 0 has ${w}: rows must be the same length (use null for an empty cell)`); });\n d.matrix.forEach((r, i) => r.forEach((x, j) => { if (!isVal(x)) bad.push(`matrix[${i}][${j}] is ${JSON.stringify(x)}, not a number or null`); }));\n if (style && style !== \"matrix\") bad.push(`--style ${style} needs stacked data (groups or series), not a matrix`);\n return bad;\n }\n const list = d.groups || d.series;\n if (!Array.isArray(list) || !list.length) return [\"the data needs a non-empty \\\"groups\\\" or \\\"series\\\" list (or a \\\"matrix\\\")\"];\n if (style === \"matrix\") bad.push(\"--style matrix needs {matrix: [[...]]}\");\n const n = Array.isArray(d.q) ? d.q.length : list[0] && Array.isArray(list[0].v) ? list[0].v.length : 0;\n list.forEach((g, i) => {\n const name = g && g.name ? `\"${g.name}\"` : `#${i}`;\n if (!g || !Array.isArray(g.v)) { bad.push(`series ${name} has no \"v\" list`); return; }\n if (g.v.length !== n) bad.push(`series ${name} has ${g.v.length} values, expected ${n}${Array.isArray(d.q) ? \" (one per period in q)\" : \"\"}`);\n g.v.forEach((x, j) => { if (!isVal(x)) bad.push(`series ${name} value ${j} is ${JSON.stringify(x)}, not a number or null`); });\n });\n if (n < 1) bad.push(\"no values to paint\");\n return bad;\n}\nconst problems = validate(data);\nif (problems.length) { console.error(`${dataPath}: ` + problems.slice(0, 10).join(\"\\n \")); process.exit(2); }\nconst palette = palArg === undefined ? undefined : palArg.includes(\",\") || palArg.startsWith(\"#\") ? palArg.split(\",\") : palArg;\nconst opts = { style, seed, size, palette, ghosts, cell: cell ? Number(cell) : undefined, maxScale: maxScale ? Number(maxScale) : undefined, background };\n\nconst pw = loadPlaywright();\nconst browser = await launch(pw);\nconst page = await browser.newPage({ viewport: { width: size[0], height: size[1] } });\nconst failed = new Promise(res => page.on(\"pageerror\", e => res(e.message)));\nawait page.addInitScript(([d, o]) => { window.DATA = d; window.OPTS = o; }, [data, opts]);\nawait page.goto(pathToFileURL(path.join(path.dirname(fileURLToPath(import.meta.url)), \"art.html\")).href);\nconst err = await Promise.race([failed, page.waitForFunction(() => document.title === \"done\", null, { timeout: 600000 }).then(() => null)]);\nif (err) { console.error(\"page error:\", err); await browser.close(); process.exit(1); }\nconst url = await page.evaluate(() => document.getElementById(\"a\").toDataURL(\"image/png\"));\nfs.writeFileSync(out, Buffer.from(url.split(\",\")[1], \"base64\"));\nawait browser.close();\nconsole.log(`wrote ${out} (${style || (data.matrix ? \"matrix\" : \"grid\")}, seed ${seed}, ${size.join(\"x\")}${(style || \"grid\") === \"grid\" && !data.matrix ? `, ghosts ${ghosts}` : \"\"}${background !== \"paper\" ? `, ${background}` : \"\"})`);\n","watercolor-data-art/scripts/preview_svg.py","#!/usr/bin/env python3\n\"\"\"A small watercolor-like SVG of stacked data, for previews where a PNG is too heavy.\n\nUsage:\n python3 preview_svg.py DATA.json OUT.svg [--seed 20261010] [--size 600x210] [--cell 28] [--layers 3]\n\nIt reads the same stacked data as render.mjs ({q?, groups|series: [{name, v, kind?, color?}],\npalette?}) and draws the grid style in miniature: series stacked from the floor with cosine\neasing, one square per grid cell inside a band, each square a few translucent layers of a\npolygon deformed by recursive midpoint displacement (Tyler Hobbs' method, without the masks, which\nSVG cannot afford at this size). No paper and no ghosts. The background is transparent, and the\ncolours are CSS custom properties: the paper pigments by default, lifted pigments (as render.mjs\n--background dark) and denser layers under prefers-color-scheme: dark. Same data and seed give the same bytes.\nStandard library only.\n\"\"\"\nimport argparse\nimport colorsys\nimport json\nimport math\n\nKIND = {\"fall\": [\"74,104,138\", \"96,128,160\", \"60,86,118\", \"118,146,168\", \"84,114,144\"],\n \"rise\": [\"168,70,50\", \"192,114,62\", \"156,90,70\", \"198,142,84\", \"146,58,46\"],\n \"steady\": [\"194,184,164\", \"172,162,144\", \"208,196,170\", \"160,154,140\", \"184,170,146\"]}\nNAMED = {\"earth\": [\"168,70,50\", \"198,142,84\", \"122,128,84\", \"74,104,138\", \"146,58,46\", \"194,184,164\", \"96,128,160\"],\n \"sea\": [\"38,70,104\", \"74,104,138\", \"96,128,160\", \"118,146,168\", \"64,124,128\", \"150,170,180\", \"50,90,110\"],\n \"dusk\": [\"92,62,102\", \"146,58,46\", \"198,142,84\", \"168,112,140\", \"74,104,138\", \"214,184,140\"],\n \"ink\": [\"40,40,44\", \"84,84,90\", \"128,126,124\", \"170,166,160\", \"60,60,70\"]}\n\n\nclass Rng:\n \"\"\"The same linear congruential generator as art.html.\"\"\"\n\n def __init__(self, seed):\n self.s = seed & 0xFFFFFFFF\n\n def __call__(self):\n self.s = (self.s * 1664525 + 1013904223) & 0xFFFFFFFF\n return self.s / 4294967296\n\n def gauss(self):\n u = v = 0.0\n while not u:\n u = self()\n while not v:\n v = self()\n return math.sqrt(-2 * math.log(u)) * math.cos(2 * math.pi * v)\n\n\ndef to_rgb(c):\n c = str(c).strip()\n if c.startswith(\"#\"):\n h = c[1:] if len(c) == 7 else \"\".join(x * 2 for x in c[1:4])\n return [int(h[i:i + 2], 16) for i in (0, 2, 4)]\n return [int(float(x)) for x in c.split(\",\")]\n\n\ndef lift(rgb):\n \"\"\"render.mjs's dark-ground pigment: lightness at least 0.62, saturation x1.15, same hue.\"\"\"\n h, l, s = colorsys.rgb_to_hls(*[v / 255 for v in rgb])\n return [round(v * 255) for v in colorsys.hls_to_rgb(h, max(l, 0.62), min(1.0, s * 1.15))]\n\n\ndef colours(D):\n \"\"\"One colour per series, by art.html's rules: own colour, else a shade of its kind, else\n the next colour of the named palette or list.\"\"\"\n lst = D.get(\"groups\") or D.get(\"series\")\n pal = D.get(\"palette\") or (\"kind\" if any(g.get(\"kind\") for g in lst) else \"earth\")\n ci, free, out = {\"fall\": 0, \"rise\": 0, \"steady\": 0}, 0, []\n for g in lst:\n k = g.get(\"kind\")\n kind = k if k in (\"rise\", \"fall\") else (\"steady\" if k else None)\n if g.get(\"color\"):\n out.append(to_rgb(g[\"color\"]))\n elif pal == \"kind\" and kind:\n out.append(to_rgb(KIND[kind][ci[kind] % 5]))\n ci[kind] += 1\n else:\n p = pal if isinstance(pal, list) else NAMED.get(\"earth\" if pal == \"kind\" else pal, NAMED[\"earth\"])\n out.append(to_rgb(p[free % len(p)]))\n free += 1\n return out\n\n\ndef deform(poly, rounds, rnd):\n for _ in range(rounds):\n out = []\n for i, (x, y, v) in enumerate(poly):\n bx, by, _ = poly[(i + 1) % len(poly)]\n s = math.hypot(bx - x, by - y) * v * 0.5\n v2 = v * (0.8 + rnd() * 0.15)\n out += [(x, y, v2), ((x + bx) / 2 + rnd.gauss() * s, (y + by) / 2 + rnd.gauss() * s, v2 * (0.8 + rnd() * 0.4))]\n poly = out\n return poly\n\n\ndef path(poly):\n \"\"\"A closed path in whole pixels, relative moves after the first point: small files.\"\"\"\n pts = [(round(x), round(y)) for x, y, _ in poly]\n d = [f\"M{pts[0][0]} {pts[0][1]}l\"]\n for (x0, y0), (x1, y1) in zip(pts, pts[1:]):\n d.append(f\"{x1 - x0} {y1 - y0}\")\n return \" \".join(d).replace(\" -\", \"-\") + \"z\"\n\n\ndef main():\n ap = argparse.ArgumentParser(description=__doc__, formatter_class=argparse.RawDescriptionHelpFormatter)\n ap.add_argument(\"data\")\n ap.add_argument(\"out\")\n ap.add_argument(\"--seed\", type=int, default=20261010)\n ap.add_argument(\"--size\", default=\"600x210\")\n ap.add_argument(\"--cell\", type=float, default=28)\n ap.add_argument(\"--layers\", type=int, default=3)\n a = ap.parse_args()\n D = json.load(open(a.data, encoding=\"utf-8\"))\n groups = D.get(\"groups\") or D.get(\"series\")\n if not groups:\n raise SystemExit(f\"{a.data}: needs stacked data, a \\\"groups\\\" or \\\"series\\\" list\")\n W, H = (int(n) for n in a.size.split(\"x\"))\n rnd = Rng(a.seed)\n cols = colours(D)\n X0, X1, FLOOR = 16, W - 16, H - 10\n Q = len(groups[0][\"v\"])\n tot = max(sum(g[\"v\"][i] or 0 for g in groups) for i in range(Q))\n S = (FLOOR - 8) / (tot or 1)\n\n def at(v, x):\n if Q == 1:\n return v[0] or 0\n f = max(0.0, min(Q - 1, (x - X0) / (X1 - X0) * (Q - 1)))\n i = min(Q - 2, int(f))\n s = (1 - math.cos((f - i) * math.pi)) / 2\n return (v[i] or 0) * (1 - s) + (v[i + 1] or 0) * s\n\n def band(x, y):\n top = FLOOR\n for k, g in enumerate(groups):\n h = at(g[\"v\"], x) * S\n if top - h < y <= top:\n return k\n top -= h\n return -1\n\n C = a.cell\n x0 = X0 + ((X1 - X0) % C) / 2\n shapes = []\n cx = x0 + C / 2\n while cx < X1:\n cy = FLOOR - C / 2\n while cy > 0:\n k = band(cx, cy)\n if k >= 0:\n h = C * 0.4\n base = deform([(x, y, 0.1 + rnd() * 0.25) for x, y in\n ((cx - h, cy - h), (cx + h, cy - h), (cx + h, cy + h), (cx - h, cy + h))], 1, rnd)\n op = 0.26 + rnd() * 0.1 # a little variation, as art.html's colour jitter\n shapes.append((k, op, [path(deform(base, 1, rnd)) for _ in range(a.layers)]))\n cy -= C\n cx += C\n\n light = \";\".join(f\"--wc{k}:rgb({','.join(map(str, c))})\" for k, c in enumerate(cols))\n dark = \";\".join(f\"--wc{k}:rgb({','.join(map(str, lift(c)))})\" for k, c in enumerate(cols))\n css = (f\".wc{{{light}}}@media (prefers-color-scheme:dark){{.wc{{{dark}}}}}\"\n + \".wc g{fill-opacity:var(--o)}@media (prefers-color-scheme:dark){.wc g{fill-opacity:calc(var(--o)*1.6)}}\"\n + \"\".join(f\".wc .s{k}{{fill:var(--wc{k})}}\" for k in range(len(cols))))\n names = \", \".join(str(g.get(\"name\", k)) for k, g in enumerate(groups))\n o = [f'\")\n svg = \"\\n\".join(o) + \"\\n\"\n open(a.out, \"w\", encoding=\"utf-8\").write(svg)\n print(f\"wrote {a.out} ({W}x{H}, {len(shapes)} squares, {a.layers} layers, seed {a.seed}, {len(svg.encode()):,} bytes)\")\n\n\nif __name__ == \"__main__\":\n main()\n","watercolor-data-art/scripts/art.html","\n\n\n

\n\n\n","watercolor-data-art/examples/readme-sections/matrix.json","{\n \"_note\": \"README section headings (#, ## and ### lines) in five public repos, counted on Dec 31 of each year. null: the repo had no README with history yet. Made from text-rule-trends/examples/readme-sections.\",\n \"rows\": [\n \"chalk/chalk\",\n \"expressjs/cors\",\n \"jshttp/mime-types\",\n \"pallets/itsdangerous\",\n \"sindresorhus/ky\"\n ],\n \"cols\": [\n \"2011\",\n \"2012\",\n \"2013\",\n \"2014\",\n \"2015\",\n \"2016\",\n \"2017\",\n \"2018\",\n \"2019\",\n \"2020\",\n \"2021\",\n \"2022\",\n \"2023\",\n \"2024\",\n \"2025\"\n ],\n \"matrix\": [\n [\n null,\n null,\n 15,\n 17,\n 15,\n 15,\n 20,\n 20,\n 19,\n 20,\n 19,\n 19,\n 19,\n 22,\n 22\n ],\n [\n null,\n null,\n 13,\n 13,\n 13,\n 13,\n 13,\n 13,\n 13,\n 13,\n 13,\n 13,\n 13,\n 13,\n 12\n ],\n [\n null,\n null,\n null,\n 11,\n 11,\n 11,\n 11,\n 11,\n 11,\n 11,\n 11,\n 11,\n 11,\n 12,\n 12\n ],\n [\n 0,\n 0,\n 0,\n 0,\n 0,\n 0,\n 0,\n 2,\n 2,\n 2,\n 2,\n 2,\n 2,\n 3,\n 4\n ],\n [\n null,\n null,\n null,\n null,\n null,\n null,\n null,\n 21,\n 24,\n 24,\n 24,\n 25,\n 25,\n 25,\n 29\n ]\n ],\n \"palette\": \"sea\"\n}\n","watercolor-data-art/examples/made-up/stack.json","{\n \"_note\": \"Made-up numbers: four series over eight periods. Two have a kind, one has its own colour (plum), and \\\"growing\\\" has neither, so it takes the first colour of the palette.\",\n \"q\": [\n \"2025-01\",\n \"2025-02\",\n \"2025-03\",\n \"2025-04\",\n \"2025-05\",\n \"2025-06\",\n \"2025-07\",\n \"2025-08\"\n ],\n \"series\": [\n {\n \"name\": \"base\",\n \"v\": [\n 2,\n 2,\n 2.2,\n 2.1,\n 2,\n 2,\n 2.1,\n 2.2\n ],\n \"kind\": \"steady\"\n },\n {\n \"name\": \"shrinking\",\n \"v\": [\n 4,\n 4.5,\n 5,\n 4,\n 3,\n 2.4,\n 2,\n 1.8\n ],\n \"kind\": \"fall\"\n },\n {\n \"name\": \"growing\",\n \"v\": [\n 1,\n 1.2,\n 1.5,\n 2,\n 2.6,\n 3.1,\n 3.4,\n 3.8\n ]\n },\n {\n \"name\": \"spiky\",\n \"v\": [\n 0.5,\n 2,\n 0.6,\n 2.2,\n 0.7,\n 2.4,\n 0.8,\n 2.5\n ],\n \"color\": \"#8c5a7a\"\n }\n ]\n}\n","watercolor-data-art/assets/preview.svg","\n","watercolor-data-art/SKILL.md","# Watercolor data art\n\nPaints data as an abstract watercolor picture, where every square or band sits where the data puts it.\n\n![Preview](assets/preview.svg)\n\nIt is not a chart: no axes, no labels. Styles: a grid of soft squares with faint ghost squares for\nwhat each series lost since its peak, Rothko-like blurred bands, random-size squares, and a matrix\nof squares sized by value. The palette comes from the data, a list of colours, a named palette, or\nthe kind of each series (rise, fall, steady). Seeded and reproducible, at any size, on warm paper,\non a transparent canvas or on a dark ground.\n\nUse it when you want an image that comes from the data rather than decoration, next to the real\nchart or with a caption that says what it shows.\n\n## Requirements\n\nNode 18+ and Playwright with Chromium. In the folder you run from:\n`npm i playwright && npx playwright install chromium`. `render.mjs` looks for Playwright in the\nworking directory, then in `NODE_PATH`, then beside the script. `PLAYWRIGHT_MODULE` chooses a\npackage folder.\n\nEach Playwright version wants its own Chromium build. A plain `npm i playwright` installs the\nnewest, which may want a build you do not have. Either run `npx playwright install chromium` with\nthat same version, pin the version that matches your browsers (e.g. `npm i playwright@1.56.1`\nfor Chromium build 1194), or set `CHROMIUM_PATH` to any installed\nChromium or Chrome. Without `CHROMIUM_PATH`, `render.mjs` also tries any Chromium Playwright\ndownloaded earlier (in `PLAYWRIGHT_BROWSERS_PATH` or `~/.cache/ms-playwright`) and says which one\nit used. Pixel-identical output needs the same Chromium build. No Python is needed (for WebP, use `tufte-charts/scripts/to_webp.py`).\n\n## Inputs\n\nA JSON file, one of:\n\n```json\n{\"q\": [\"2025Q2\", \"...\"], \"groups\": [{\"name\": \"...\", \"v\": [1.4, \"...\"], \"kind\": \"rise\", \"color\": \"#a33a2c\"}]}\n{\"series\": [{\"name\": \"...\", \"v\": [...]}], \"palette\": \"sea\"}\n{\"matrix\": [[3, 5, null], [1, 0, 2]], \"rows\": [\"...\"], \"cols\": [\"...\"], \"palette\": [\"#264668\", \"#4a688a\"], \"diverging\": false}\n```\n\n- `render.mjs` checks the input before it starts a browser: every value must be a number or\n `null`, and every series must have one value per period (per entry in `q`, if given). Otherwise\n it exits with status 2 and says which series and value are wrong.\n- Stacked data (`groups` or `series`, bottom to top) is what `text-rule-trends` writes as\n `themes.json`. `v` is one value per period. `kind` and `color` are optional.\n- A matrix is rows of values, all rows the same length. `null` is an empty cell, painted as a ghost. 0 is left blank.\n Square sides grow with the square root of |value|. Colour is by row, or by sign with `\"diverging\": true`.\n\n**Palette.** A named `--palette` (or a list) overrides the colours by kind: kinds no longer pick\ncolours, and every series without its own `color` takes the palette in turn.\n`--palette` or the data's `palette` is `kind`, a named palette (`earth`, `sea`,\n`dusk`, `ink`), or a list of colours (`#hex` or `r,g,b`). The default is `kind` when any series\nhas a kind, else `earth`. Each series gets, in order:\n1. its own `color`, if it has one;\n2. with `kind`: the next shade of its kind (warm earth for rise, cool ink for fall, sand for\n steady or too few; 5 shades each);\n3. otherwise the next colour of the named palette or list (`earth` when the palette is `kind`),\n counted over these series only. So a series with no kind and no colour never repeats a colour\n that another series took by kind or by its own `color`, until the palette runs out.\n\n## Steps\n\nWrite outputs to a folder of your own. `SKILL` is this skill's folder.\n\n```bash\nnode $SKILL/scripts/render.mjs data.json grid.png # grid, the default\nnode $SKILL/scripts/render.mjs data.json rothko.png --style rothko\nnode $SKILL/scripts/render.mjs data.json squares.png --style squares --seed 7 --size 1200x420\nnode $SKILL/scripts/render.mjs matrix.json matrix.png --size 1500x560 # matrix data picks the matrix style\n```\n\nOptions: `--style grid|rothko|squares|matrix`, `--seed N` (default 20261010), `--size WxH`\n(default 2400x840), `--palette ...`, `--ghosts running|global|none` (grid only, default `running`),\n`--cell PX` (square size; default 26 for grid, fitted for matrix), `--max-scale N`,\n`--background paper|transparent|dark` (default `paper`).\n\n## Light and dark\n\n- `--background paper` (default): warm paper `#f7f4ec` with a faint grain, and the pigments as\n given. This is how every image before this option was made, pixel for pixel.\n- `--background transparent`: no paper and no grain; the canvas stays clear, so the PNG has an\n alpha channel and sits on any page. The pigments are lifted for a dark ground: each colour keeps\n its hue, its lightness is raised to at least 0.62 and its saturation by 15%. On a white page\n they look pastel, so use `paper` for light pages.\n- `--background dark`: the same lifted pigments on a flat `#111` ground, no grain.\n- All three draw the same squares, edges and jitter for the same seed (the grain's random numbers\n are drawn even when no grain is painted), so a light and a dark image line up shape for shape.\n\nPair them on a website with `

\n ```\n\n Or use the SVG itself (`

`), which switches on its own; it falls back to\n Palatino or Georgia, because an SVG shown as an image cannot load the web font.\n\n## Specs\n\nEvery spec has `\"data\"`: an object, or the name of a JSON file beside the spec. Periods can be\nquarters (`2026Q1`), months (`2026-03`), years (`2026`), days (`2026-03-15`) or any labels; dated\nperiods sit on a true time scale (and allow `events` and `quote`), other labels are evenly spaced.\nEvery spec may also set `\"background\"`: `\"transparent\"` (the default, light and dark) or\n`\"paper\"` (the same as `--paper`). Each script's `--help` lists every key. In short:\n\n| script | data | main keys |\n|---|---|---|\n| `draw_stack.py` | `{q, groups: [{name, v, kind?}], tot?, nfiles?, from_q?, to_q?, rest?}`, bottom to top | `title`, `subtitle`, `key` (`[red]..[/]`, `[blue]..[/]`), `ylabel`, `share_label`, `count_label`, `arrow`, `quote`, `events`, `footnote`, `threshold`, `y_step`, `baseline` (`zero` or `center`), `layout` (`plot_height` 560: the stack's top is fitted to it; `max_scale`: optional cap in px per unit) |\n| `draw_lines.py` (y ticks and values print as plain numbers by default; set `y.tick` and `value_format` to `\"{:.0f}%\"` for shares) | anything; each series has `values`, or `from` a key (`{\"key\": \"tot\"}` reads text-rule-trends' `pooled.json` directly; add `\"sample\"` for share.json), or `from` a share of counts | `periods`, `series [{label, color, ...}]`, `y {min, max, step, tick, reference}`, `value_format`, `count_label`, `events`, `quote`, `footnote` |\n| `draw_slope.py` | `{items: [{name, from, to}]}`, or stacked data plus `left`, `right`, `measure` | `title`, `left_label`, `right_label`, `value_format`, `threshold`, `footnote` |\n| `draw_multiples.py` | `{q, series: [{name, v}]}`, or stacked data | `columns`, `panel`, `y`, `value_format`, `order`, `threshold`, `footnote` |\n\nPlaceholders: the stack chart fills `total_from`, `total_to`, `change_pct`, `p_from`, `p_to`,\n`n_total` (and the older `tot_from`, `tot_to`, `q_from`, `q_to`, `files`). The line chart fills\n`{a[i]}`, `{b[i]}`, ... (each series' values, i over the periods drawn) and `{a_last}` (the last\nvalue that is not null). Both fill `{data[...]}` from the data file, e.g.\n`{data[samples][][files]:,}` or `{data[files_with_items]:,}` from text-rule-trends'\nthemes.json, so counts in a footnote come from the data too. Use Python format specs:\n`{change_pct:+.0f}`.\n\nLayout details the scripts handle: if either compared bin is empty (or its shares are null), the\nstack chart draws no arrow, shows those shares as \"–\" and colours the bands as too few. A rising\narrow gets headroom on the y axis. With no `events` and no `quote`, the stack and line charts close\nup the row kept for them (set `layout.keep_event_row` to keep it). A missing spec or data file\nstops with its path, and a placeholder the chart does not have is named with the list of those it\nhas. Missing values (an empty bin) print as \"–\". The slopegraph stops with a message if a compared\nbin is empty. A value label at the floor of a line chart goes above its point, off the x axis. The stack chart measures its label column, and puts each share\non a line under its name (wrapping very long names) when a name would run into its share. It\nlifts the arrow's words clear of the bands. Empty bins (0 files or null values) are skipped. The\nline chart prints counts only for series that have them. The slopegraph draws a thin leader from\neach end of a line to its label, so tied values pushed apart stay readable. Small multiples sit on\na true time scale with a faint tick per period.\n\nA group with no `kind` gets one from its share of the total between `from_q` and `to_q`. `too few`\n(from `text-rule-trends`, when a compared bin is under 20 files) is drawn grey like `steady`.\n\n## Examples\n\n### README sections, from public data (`examples/readme-sections/`)\n\n`themes.data.json` is the output of `text-rule-trends/examples/readme-sections`.\n`stack.spec.json`, `slope.spec.json` and `multiples.spec.json` draw it three ways: headings per\nREADME rose from 13.0 to 15.8 (2016 to 2025), with \"API reference entries\" the one theme that grew.\n\n### Made-up lines with gaps (`examples/made-up/lines.spec.json`)\n\nMonthly values with nulls, a negative value and a null last value. The lines break at the gaps,\nthe label sits at each series' last known value, and −3 prints with a true minus.\n\n### Smoke test\n\n```bash\npython3 $SKILL/scripts/draw_stack.py --spec $SKILL/examples/readme-sections/stack.spec.json --out /tmp/tc/stack.html\npython3 $SKILL/scripts/draw_lines.py --spec $SKILL/examples/made-up/lines.spec.json --out /tmp/tc/lines.html\nnode $SKILL/scripts/render.mjs /tmp/tc/stack.html /tmp/tc/stack.png # stack.light.png, stack.dark.png\n```\n\n`draw_stack.py` prints `stack top 15.80, 31.1px per unit, 7 bins with data`; `draw_lines.py` prints\n`series A 42, 48, -, 61, −3, -; series B 30, 33, 35, -, 41, 44`. The two PNGs are 2400 px wide,\nwith transparent backgrounds.\n\n### The preview (`assets/preview.svg`)\n\n`examples/readme-sections/preview.spec.json` draws the README example at 600x290 with no title:\n`draw_stack.py --spec $SKILL/examples/readme-sections/preview.spec.json --out preview.html --svg preview.svg`.\n\n**Worked example.** These skills were first written for the agentleFS article \"CLAUDE.md best\npractices\" (https://agentlefs.com/blog/claude-md-best-practices), which studied 5,786 CLAUDE.md and\nAGENTS.md files. Its data, settings and step-by-step recipe live outside the skills, in the\n`article-claude-md-best-practices` folder (`RECIPE.md` and `data/`). Its figures use the classic\npaper look, so the recipe passes `--paper`.\n\n## Limits and caveats\n\n- Layouts are tuned for about 12 bands, 6 to 8 bins and labels up to about 40 characters at\n 1200 px wide. Change `layout` in the spec for other shapes.\n- The slopegraph and the small multiples size themselves to the items; very long names need a\n wider `layout`.\n- Fonts come from a CDN. Offline rendering changes the face and text widths slightly.\n- PNGs are identical only on the same Chromium build; fonts and antialiasing differ across builds.\n- A transparent PNG's band separators and label halos are drawn in `--tc-gap` (`#fffff8` or\n `#111`), so on a page of a very different colour they show as faint lines. Use the SVG inline\n and set `--tc-gap`, or render with `--paper`.\n\n## Credits\n\n- **Edward Tufte**, *The Visual Display of Quantitative Information* (Graphics Press, 1983; 2nd\n ed. 2001). The principles only: maximise data-ink, no chart junk, range frames, direct labels,\n small multiples and the slopegraph. No text or figure is copied from the book.\n- **ET Book**, the typeface, by Dmitry Krasny, Bonnie Scranton and Edward Tufte (MIT licence),\n loaded from the tufte-css 1.8.0 package by Dave Liepmann (MIT licence) on jsDelivr. It is\n linked, not bundled.\n- **F. N. Fritsch and R. E. Carlson**, \"Monotone Piecewise Cubic Interpolation\", SIAM Journal on\n Numerical Analysis 17(2), 1980: the slopes that keep the curve between bins from overshooting.\n- Playwright (Apache-2.0) and Pillow (MIT-CMU) are used as installed tools, not bundled.\n","tufte-charts/LICENSE","text-rule-trends/scripts/extract.py","#!/usr/bin/env python3\n\"\"\"Extract line-level items (by default \"rules\") from every version of every file, and follow\neach item's life: when it appeared, and when it was deleted.\n\nReads SAMPLE/vers/*.tsv.gz or *.tsv as written by github-file-history (one row per version:\nsha, author_date, claude_coauthor 0/1, base64 body; an empty body is a deletion).\n\nUsage:\n python3 extract.py SAMPLE_DIR [--extractor rules|lines|regex|plugin] [--regex RE]\n [--plugin FILE.py:FUNC] [--min-words N] [--max-words N] [--match 0.6]\n [--order datetime|string] [--out FILE]\n\nExtractors:\n rules (default) a list item, or a line starting with a capital letter, whose text (after an\n optional bold label) starts with an imperative or normative word (always, never, use,\n run, avoid, must, should, ...). Fenced code, headings and table rows are skipped.\n Default words: 4 to 45.\n lines every non-blank line outside fenced code, list markers removed. Default words: 1 to 45.\n regex lines (outside fenced code) that match --regex. The item is group 1 if the pattern has\n a group, else the whole match. Example, Markdown headings: --regex '^#{1,3}\\\\s+(.+)'\n plugin your own function: --plugin myextract.py:items. It gets the file body (str) and returns\n a list of item strings. Word limits and the identity rules below still apply.\n\nWords are counted after normalising: lower case, code spans -> \"code\", links -> their text,\ndigits and markdown removed. An item keeps its identity across a rewording when most of its\nwords survive: token Jaccard >= --match (default 0.6) against an unmatched item of the\nprevious version.\n\nVersions are replayed oldest first. --order datetime (default) sorts by the parsed author date,\nso mixed UTC offsets order correctly. --order string sorts by the date text, as some older\nruns did.\n\nOutput (default SAMPLE_DIR/rules.json):\n nfiles files with at least one version\n nhistory history files found (a file whose history came back empty counts here, not in nfiles)\n life one record per item: {kind, born, died, text, norm, repo}\n kind the file's kind (its file name unless --kind-map says otherwise)\n born author date of the version where it first appeared\n died author date of the version where it disappeared, or null\n text the line as written; norm its normalised form\n repo the history file's name (.tsv.gz or .tsv), i.e. one file in one\n repo; the key is ___. Kept as \"repo\" for compatibility\n files {key: {kind, first, deleted}}: first version date, and the date of the deletion if\n the last version deleted the file\n co_month {\"YYYY-MM\": [versions with a Claude co-author trailer, versions]}\n settings the extractor settings used\n\"\"\"\nimport argparse\nimport collections\nimport importlib.util\nimport json\nimport os\nimport re\nimport sys\nfrom datetime import datetime\n\nsys.path.insert(0, os.path.dirname(os.path.abspath(__file__)))\nfrom common import Kinds, history_files, key_of, load_kind_map, load_versions # noqa: E402\n\nWORD = re.compile(r\"[a-z][a-z0-9_+#.-]*\")\nIMPERATIVE = re.compile(r\"^(always|never|do|don'?t|does not|avoid|prefer|use|run|keep|make sure|ensure|follow|add|write|read|check|test|update|create|remove|delete|put|place|import|call|ask|verify|commit|document|must|should|only|before|after|when|if|no |not |all |every|treat|leave|stop|start|wait|search|look|consider|default|instead|don't)\\b\")\n\n\ndef norm(line):\n t = re.sub(r\"`[^`]*`\", \" code \", line.lower())\n t = re.sub(r\"\\[([^\\]]*)\\]\$[^)]*\$\", r\"\\1\", t)\n t = re.sub(r\"[*_>#]\", \" \", t)\n t = re.sub(r\"\\d+\", \" \", t)\n return \" \".join(WORD.findall(t))\n\n\ndef rule_lines(body):\n \"\"\"The default heuristic: imperative list items and capitalised lines.\"\"\"\n out, fence = [], False\n for raw in body.splitlines():\n s = raw.strip()\n if s.startswith(\"```\"):\n fence = not fence\n continue\n if fence or not s or s.startswith(\"#\") or s.startswith(\"|\"):\n continue\n m = re.match(r\"^([-*+]|\\d+[.)])\\s+(.*)\", s)\n text = m.group(2) if m else s\n if not m and not text[:1].isupper():\n continue # a wrapped continuation line, not a rule\n head = re.sub(r\"^\\*\\*[^*]{1,60}\\*\\*\\s*[:.-]?\\s*\", \"\", text) # drop a bold label\n if IMPERATIVE.match(head.lower()):\n out.append(text)\n return out\n\n\ndef plain_lines(body, rx=None):\n out, fence = [], False\n for raw in body.splitlines():\n s = raw.strip()\n if s.startswith(\"```\"):\n fence = not fence\n continue\n if fence or not s:\n continue\n if rx is None:\n m = re.match(r\"^([-*+]|\\d+[.)])\\s+(.*)\", s)\n out.append(m.group(2) if m else s)\n else:\n m = rx.search(s)\n if m:\n out.append(m.group(1) if m.groups() else m.group(0))\n return out\n\n\ndef load_plugin(spec):\n path, _, func = spec.rpartition(\":\")\n if not path:\n raise SystemExit(\"--plugin wants FILE.py:FUNCTION\")\n mod_spec = importlib.util.spec_from_file_location(\"extract_plugin\", path)\n mod = importlib.util.module_from_spec(mod_spec)\n mod_spec.loader.exec_module(mod)\n return getattr(mod, func)\n\n\ndef make_items(a):\n if a.extractor == \"rules\":\n get = rule_lines\n elif a.extractor == \"lines\":\n get = plain_lines\n elif a.extractor == \"regex\":\n if not a.regex:\n raise SystemExit(\"--extractor regex needs --regex\")\n rx = re.compile(a.regex)\n get = lambda body: plain_lines(body, rx)\n else:\n if not a.plugin:\n raise SystemExit(\"--extractor plugin needs --plugin FILE.py:FUNC\")\n get = load_plugin(a.plugin)\n lo = a.min_words if a.min_words is not None else (4 if a.extractor == \"rules\" else 1)\n hi = a.max_words\n\n def items(body):\n seen, uniq = set(), []\n for text in get(body):\n n = norm(text)\n if lo <= len(n.split()) <= hi and n not in seen:\n seen.add(n)\n uniq.append((n, text.strip()))\n return uniq\n return items, lo\n\n\ndef jacc(a, b):\n A, B = set(a.split()), set(b.split())\n return len(A & B) / max(1, len(A | B))\n\n\ndef parse_dt(s):\n try:\n return datetime.fromisoformat(s.replace(\"Z\", \"+00:00\")).timestamp()\n except ValueError:\n return float(\"inf\")\n\n\ndef main():\n ap = argparse.ArgumentParser(description=__doc__, formatter_class=argparse.RawDescriptionHelpFormatter)\n ap.add_argument(\"sample\")\n ap.add_argument(\"--extractor\", default=\"rules\", choices=[\"rules\", \"lines\", \"regex\", \"plugin\"])\n ap.add_argument(\"--regex\")\n ap.add_argument(\"--plugin\")\n ap.add_argument(\"--min-words\", type=int)\n ap.add_argument(\"--max-words\", type=int, default=45)\n ap.add_argument(\"--match\", type=float, default=0.6, help=\"Jaccard needed to keep identity through a rewording\")\n ap.add_argument(\"--order\", default=\"datetime\", choices=[\"datetime\", \"string\"])\n ap.add_argument(\"--kind-map\", help='JSON file or string {\"regex\": \"label\"} matched against the path')\n ap.add_argument(\"--out\", help=\"default SAMPLE_DIR/rules.json\")\n a = ap.parse_args()\n\n items_of, lo = make_items(a)\n kinds = Kinds(a.sample, load_kind_map(a.kind_map))\n files = history_files(a.sample)\n if not files:\n sys.exit(f\"no {a.sample}/vers/*.tsv.gz or *.tsv\")\n life, nfiles, info = [], 0, {}\n co_month = collections.defaultdict(lambda: [0, 0])\n for path in files:\n key = key_of(path)\n kind = kinds(key)\n vers = load_versions(path)\n if a.order == \"datetime\":\n vers.sort(key=lambda r: parse_dt(r[0]))\n if not vers:\n continue\n nfiles += 1\n info[key] = {\"kind\": kind, \"first\": vers[0][0], \"deleted\": vers[-1][0] if not vers[-1][2] else None}\n alive = {} # norm -> record\n for d, co, body in vers:\n co_month[d[:7]][0] += co\n co_month[d[:7]][1] += 1\n matched, nxt = set(), {}\n for n, text in items_of(body):\n rec = alive.get(n)\n if rec is None:\n best = max(((jacc(n, k), k) for k in alive if k not in matched), default=(0, None))\n if best[0] >= a.match:\n rec = alive[best[1]]\n matched.add(best[1])\n if rec is None:\n rec = {\"kind\": kind, \"born\": d, \"died\": None, \"text\": text, \"norm\": n, \"repo\": os.path.basename(path)}\n life.append(rec)\n else:\n matched.add(rec[\"norm\"])\n rec[\"norm\"] = n\n nxt[n] = rec\n kept = {id(r) for r in nxt.values()}\n for rec in alive.values():\n if id(rec) not in kept:\n rec[\"died\"] = d\n alive = nxt\n settings = {\"extractor\": a.extractor, \"regex\": a.regex, \"plugin\": a.plugin, \"min_words\": lo,\n \"max_words\": a.max_words, \"match\": a.match, \"order\": a.order}\n out = a.out or os.path.join(a.sample, \"rules.json\")\n with open(out, \"w\") as f:\n json.dump({\"nfiles\": nfiles, \"nhistory\": len(files), \"life\": life, \"files\": info, \"co_month\": co_month, \"settings\": settings}, f)\n with_items = len({r[\"repo\"] for r in life})\n gone = sum(r[\"died\"] is not None for r in life)\n print(f\"files {nfiles} with items {with_items} items {len(life)} later deleted {gone / max(1, len(life)):.0%} -> {out}\")\n\n\nif __name__ == \"__main__\":\n main()\n","text-rule-trends/scripts/embed.py","#!/usr/bin/env python3\n\"\"\"Embed every item's text with a sentence model (bge-small by default).\n\nUsage:\n python3 embed.py SAMPLE_DIR [--model BAAI/bge-small-en-v1.5]\n [--backend fastembed|sentence-transformers] [--cache-dir DIR] [--max-chars 300]\n\nReads SAMPLE_DIR/rules.json, writes SAMPLE_DIR/emb.npy (float32, one row per item, same order).\nemb.npy is a local working file; it is not meant to be shared or bundled.\nBefore embedding: code spans become the word CODE, links keep their text, bold markers go, and\nthe text is cut at --max-chars characters.\n\nBackends:\n fastembed (default). \"BAAI/bge-small-en-v1.5\" is served as\n Qdrant's quantised ONNX export. Any model fastembed lists works. pip install fastembed\n sentence-transformers: any Hugging Face sentence model, normalised. pip install sentence-transformers\n Vectors differ slightly from fastembed's for the same model, so clusters will move.\n\"\"\"\nimport argparse\nimport json\nimport os\nimport re\n\nimport numpy as np\n\n\ndef clean(t, n):\n t = re.sub(r\"`[^`]*`\", \"CODE\", t)\n t = re.sub(r\"\\[([^\\]]*)\\]\$[^)]*\$\", r\"\\1\", t)\n t = re.sub(r\"\\*\\*([^*]+)\\*\\*\", r\"\\1\", t)\n return t.strip()[:n]\n\n\ndef main():\n ap = argparse.ArgumentParser(description=__doc__, formatter_class=argparse.RawDescriptionHelpFormatter)\n ap.add_argument(\"sample\")\n ap.add_argument(\"--model\", default=\"BAAI/bge-small-en-v1.5\")\n ap.add_argument(\"--backend\", default=\"fastembed\", choices=[\"fastembed\", \"sentence-transformers\"])\n ap.add_argument(\"--cache-dir\", default=None, help=\"where the model is downloaded\")\n ap.add_argument(\"--batch\", type=int, default=256)\n ap.add_argument(\"--max-chars\", type=int, default=300)\n a = ap.parse_args()\n\n life = json.load(open(os.path.join(a.sample, \"rules.json\")))[\"life\"]\n texts = [clean(r[\"text\"], a.max_chars) for r in life]\n if a.backend == \"fastembed\":\n from fastembed import TextEmbedding\n model = TextEmbedding(a.model, cache_dir=a.cache_dir)\n E = np.array(list(model.embed(texts, batch_size=a.batch)), dtype=np.float32)\n else:\n from sentence_transformers import SentenceTransformer\n model = SentenceTransformer(a.model, cache_folder=a.cache_dir)\n E = model.encode(texts, batch_size=a.batch, normalize_embeddings=True, show_progress_bar=False).astype(np.float32)\n np.save(os.path.join(a.sample, \"emb.npy\"), E)\n print(f\"embedded {len(texts)} items with {a.model} -> {E.shape}\")\n\n\nif __name__ == \"__main__\":\n main()\n","text-rule-trends/scripts/common.py","\"\"\"Shared pieces: reading file histories, file kinds, periods, centroids.\"\"\"\nimport base64\nimport gzip\nimport glob\nimport json\nimport os\nimport re\nfrom datetime import date, timedelta\n\n\ndef history_files(root):\n \"\"\"SAMPLE/vers/*.tsv.gz and *.tsv, sorted by key.\"\"\"\n files = glob.glob(os.path.join(root, \"vers\", \"*.tsv.gz\")) + glob.glob(os.path.join(root, \"vers\", \"*.tsv\"))\n return sorted(files, key=lambda p: os.path.basename(p))\n\n\ndef key_of(path):\n b = os.path.basename(path)\n return b[: -len(\".tsv.gz\")] if b.endswith(\".tsv.gz\") else b[: -len(\".tsv\")]\n\n\ndef open_text(path):\n return gzip.open(path, \"rt\") if path.endswith(\".gz\") else open(path, encoding=\"utf-8\")\n\n\ndef load_versions(path):\n \"\"\"[(author_date, claude_coauthor, body)], oldest first. An empty body is a deletion.\"\"\"\n rows = []\n with open_text(path) as f:\n for line in f:\n parts = line.rstrip(\"\\n\").split(\"\\t\")\n if len(parts) < 3:\n continue\n body = base64.b64decode(parts[3]).decode(\"utf-8\", \"replace\") if len(parts) > 3 and parts[3] else \"\"\n rows.append((parts[1], int(parts[2]), body))\n rows.sort(key=lambda r: r[0]) # oldest first, by author date string\n return rows\n\n\ndef load_manifest(root):\n \"\"\"{key: path in repo} from SAMPLE/manifest.tsv, or {} when there is none.\"\"\"\n p = os.path.join(root, \"manifest.tsv\")\n out = {}\n if os.path.exists(p):\n for line in open(p, encoding=\"utf-8\"):\n f = line.rstrip(\"\\n\").split(\"\\t\")\n if len(f) >= 3:\n out[f[0]] = f[2]\n return out\n\n\nclass Kinds:\n \"\"\"A file's kind. By default its file name (CLAUDE.md, README.md, package.json).\n\n kind_map is {regex: label}, tried in order against the file's path in the repo\n (or, with no manifest, the part of the key after \"__\"); the first match wins.\n \"\"\"\n\n def __init__(self, root, kind_map=None):\n self.manifest = load_manifest(root)\n self.rules = [(re.compile(k), v) for k, v in (kind_map or {}).items()]\n\n def path(self, key):\n return self.manifest.get(key) or key.split(\"__\", 1)[-1]\n\n def __call__(self, key):\n p = self.path(key)\n for rx, label in self.rules:\n if rx.search(p):\n return label\n return p.rsplit(\"/\", 1)[-1]\n\n\ndef load_kind_map(s):\n if not s:\n return None\n if isinstance(s, dict):\n return s\n return json.load(open(s)) if os.path.exists(s) else json.loads(s)\n\n\n# ---- periods: \"2026Q1\" (quarter), \"2026-03\" (month), \"2026\" (year)\n\ndef period_of(d, unit):\n if unit == \"quarter\":\n return f\"{d[:4]}Q{(int(d[5:7]) - 1) // 3 + 1}\"\n if unit == \"month\":\n return d[:7]\n if unit == \"year\":\n return d[:4]\n raise SystemExit(f\"unknown period {unit!r}: use quarter, month or year\")\n\n\ndef period_start(p):\n if \"Q\" in p:\n return date(int(p[:4]), 3 * int(p[5]) - 2, 1)\n if len(p) == 7:\n return date(int(p[:4]), int(p[5:7]), 1)\n return date(int(p[:4]), 1, 1)\n\n\ndef period_next(p):\n \"\"\"The first day after the period.\"\"\"\n s = period_start(p)\n if \"Q\" in p:\n return date(s.year + 1, 1, 1) if s.month == 10 else date(s.year, s.month + 3, 1)\n if len(p) == 7:\n return date(s.year + 1, 1, 1) if s.month == 12 else date(s.year, s.month + 1, 1)\n return date(s.year + 1, 1, 1)\n\n\ndef period_range(p0, p1, unit):\n out, p = [], p0\n while p <= p1:\n out.append(p)\n p = period_of(period_next(p).isoformat(), unit)\n return out\n\n\ndef period_end(p):\n return period_next(p) - timedelta(days=1)\n\n\n# ---- centroids: JSON (text) or .npy\n\ndef load_centroids(path):\n import numpy as np\n if path.endswith(\".npy\"):\n return np.load(path).astype(np.float32)\n d = json.load(open(path))\n C = d[\"centroids\"] if isinstance(d, dict) else d\n return np.array(C, dtype=np.float32)\n\n\ndef centroids_id(C):\n \"\"\"A short fingerprint of a set of centroids (sha256 of the float32 values), so a labels file\n and a themes file can be checked against each other.\"\"\"\n import hashlib\n import numpy as np\n return \"sha256:\" + hashlib.sha256(np.ascontiguousarray(np.asarray(C, dtype=np.float32)).tobytes()).hexdigest()[:16]\n\n\ndef save_centroids(path, C, **meta):\n \"\"\"Each value is written as the shortest decimal that reads back as the same float32.\"\"\"\n import numpy as np\n C = np.asarray(C, dtype=np.float32)\n rows = \",\\n \".join(\"[\" + \",\".join(np.format_float_positional(x, unique=True, trim=\"-\") for x in row) + \"]\" for row in C)\n head = json.dumps({**meta, \"id\": centroids_id(C), \"k\": int(C.shape[0]), \"dim\": int(C.shape[1])})[1:-1]\n with open(path, \"w\") as f:\n f.write(\"{\" + head + ',\\n \"centroids\": [\\n ' + rows + \"\\n ]\\n}\\n\")\n","text-rule-trends/scripts/cluster.py","#!/usr/bin/env python3\n\"\"\"Group items by meaning: k-means on one sample, nearest centroid for any other.\n\nFit (the reference sample):\n python3 cluster.py fit SAMPLE_DIR --k 40 [--n-init 6] [--seed 0] [--show 7] [--centroids-out FILE]\n [--match-to OLD_centroids.json]\n -> SAMPLE_DIR/labels.json {\"k\", \"labels\", \"centroids_id\"} and SAMPLE_DIR/centroids.json\n Prints each cluster with its size, reach and the items nearest its centre, so a person can\n name the clusters and group them into themes (see examples/*/themes.json).\n --match-to renumbers the new clusters to line up with older centroids (Hungarian matching on\n cosine similarity) and prints how close each pair is. Even then the new clusters are not the\n old ones: a themes file names one set of centroids, and analyze.py refuses a themes file\n whose centroids_id differs from the labels'. To reuse old themes on new data, assign.\n\nAssign (any other sample, to saved centroids):\n python3 cluster.py assign SAMPLE_DIR --centroids centroids.json\n -> SAMPLE_DIR/labels.json\n\nConvert saved centroids between .npy and JSON:\n python3 cluster.py convert IN.npy OUT.json (or IN.json OUT.npy)\n\nk-means is sklearn's KMeans with k-means++ init. The saved centroids are the mean embedding of\neach cluster's items, unit-normalised. Assignment picks the highest cosine similarity.\ncentroids.json is plain text: each value is the shortest decimal that reads back as the same\nfloat32, so a JSON round trip gives exactly the same labels as the .npy.\n\"\"\"\nimport argparse\nimport json\nimport os\nimport sys\n\nimport numpy as np\n\nsys.path.insert(0, os.path.dirname(os.path.abspath(__file__)))\nfrom common import centroids_id, load_centroids, save_centroids # noqa: E402\n\n\ndef show(life, E, labels, centres, n):\n for c in np.argsort(-np.bincount(labels, minlength=len(centres))):\n idx = np.where(labels == c)[0]\n if not len(idx):\n continue\n files = len({life[i][\"repo\"] for i in idx})\n alive = np.mean([life[i][\"died\"] is None for i in idx])\n print(f\"\\n#{c} {len(idx)} items in {files} files, still there {alive:.0%}\")\n for i in idx[np.argsort(-(E[idx] @ centres[c]))[:n]]:\n print(\" -\", life[i][\"text\"][:110])\n\n\ndef main():\n ap = argparse.ArgumentParser(description=__doc__, formatter_class=argparse.RawDescriptionHelpFormatter)\n ap.add_argument(\"mode\", choices=[\"fit\", \"assign\", \"convert\"])\n ap.add_argument(\"sample\", help=\"SAMPLE_DIR (convert: the input file)\")\n ap.add_argument(\"target\", nargs=\"?\", help=\"convert: the output file\")\n ap.add_argument(\"--k\", type=int, default=40)\n ap.add_argument(\"--n-init\", type=int, default=6)\n ap.add_argument(\"--seed\", type=int, default=0)\n ap.add_argument(\"--centroids\", help=\"assign: centroids.json or .npy\")\n ap.add_argument(\"--centroids-out\", help=\"fit: where to write the centroids (default SAMPLE_DIR/centroids.json)\")\n ap.add_argument(\"--show\", type=int, default=7, help=\"fit: items to print per cluster (0 = none)\")\n ap.add_argument(\"--match-to\", help=\"fit: renumber clusters to match these centroids (Hungarian on cosine)\")\n a = ap.parse_args()\n\n if a.mode == \"convert\":\n C = load_centroids(a.sample)\n if a.target.endswith(\".npy\"):\n np.save(a.target, C)\n else:\n save_centroids(a.target, C)\n back = load_centroids(a.target)\n print(f\"wrote {a.target}: {C.shape}, round trip exact: {bool(np.array_equal(back, C))}\")\n return\n\n life = json.load(open(os.path.join(a.sample, \"rules.json\")))[\"life\"]\n E = np.load(os.path.join(a.sample, \"emb.npy\"))\n if len(E) != len(life):\n raise SystemExit(\"emb.npy does not match rules.json; run embed.py again\")\n\n if a.mode == \"fit\":\n from sklearn.cluster import KMeans\n if len(E) < a.k:\n raise SystemExit(f\"{len(E)} items cannot make {a.k} clusters; lower --k\")\n km = KMeans(a.k, n_init=a.n_init, random_state=a.seed).fit(E)\n labels = km.labels_\n C = np.stack([E[labels == k].mean(0) for k in range(a.k)])\n C /= np.linalg.norm(C, axis=1, keepdims=True)\n centres = km.cluster_centers_\n if a.match_to:\n from scipy.optimize import linear_sum_assignment\n old = load_centroids(a.match_to)\n if old.shape != C.shape:\n raise SystemExit(f\"--match-to has shape {old.shape}, the fit {C.shape}\")\n sim = old @ C.T # [old, new]\n rows, cols = linear_sum_assignment(-sim)\n perm = np.empty(a.k, dtype=int)\n perm[cols] = rows # new cluster cols[i] becomes old id rows[i]\n labels = perm[labels]\n order = np.argsort(perm)\n C, centres = C[order], centres[order]\n s = sim[rows, cols]\n same = int(np.sum(np.abs(old - C).max(1) == 0))\n print(f\"matched to {a.match_to}: cosine per pair min {s.min():.3f}, median {np.median(s):.3f}; \"\n f\"{same} of {a.k} centroids identical\", flush=True)\n if same < a.k:\n print(\"warning: the new clusters are not the old ones. A themes file written for the old \"\n \"centroids will be refused by analyze.py; assign to the old centroids instead, or \"\n \"rename the clusters for this fit.\", flush=True)\n out = a.centroids_out or os.path.join(a.sample, \"centroids.json\")\n if out.endswith(\".npy\"):\n np.save(out, C.astype(np.float32))\n else:\n save_centroids(out, C, n_init=a.n_init, seed=a.seed)\n k = a.k\n cid = centroids_id(C.astype(np.float32))\n else:\n if not a.centroids:\n raise SystemExit(\"assign needs --centroids\")\n C = load_centroids(a.centroids)\n labels = (E / np.linalg.norm(E, axis=1, keepdims=True) @ C.T).argmax(1)\n k = len(C)\n cid = centroids_id(C)\n json.dump({\"k\": k, \"labels\": [int(x) for x in labels], \"centroids_id\": cid}, open(os.path.join(a.sample, \"labels.json\"), \"w\"))\n print(f\"{a.mode}: {len(labels)} items in {k} clusters -> {a.sample}/labels.json\", flush=True)\n if a.mode == \"fit\" and a.show:\n show(life, E, labels, centres, a.show)\n\n\nif __name__ == \"__main__\":\n main()\n","text-rule-trends/scripts/analyze.py","#!/usr/bin/env python3\n\"\"\"Turn item lives into trends: read each file at an age, bin by period, pool samples, map\nclusters to themes, label rise/fall/steady, compare before/after a date, and count the share\nof one file kind among new files.\n\nUsage:\n python3 analyze.py --sample NAME=DIR [--sample NAME=DIR ...] --out OUT\n [--config config.json] [--themes themes.json] [options]\n\nEvery option can also be set in --config (a JSON object whose keys are the option names with\nunderscores, e.g. {\"age\": 30, \"period\": \"quarter\", \"from\": \"2025Q4\"}). The command line wins.\n\nOptions (defaults in brackets):\n --mode cohort|snapshot [cohort] cohort: bin files by the period they were created in, and read\n each file as it stood --age days after creation. snapshot: for every\n period, read every file that existed as it stood on the period's last\n day (deleted files drop out).\n --age N [30] days after creation (cohort mode)\n --created first-item|first-version [first-item] a file's creation date: the birth of its\n earliest item, or its first commit\n --period quarter|month|year [quarter]\n --periods P1,P2,... [every period with data] the bins to report, e.g. 2025Q2,2025Q3\n --from P --to P [first and last bin] the two bins compared for rise/fall\n --threshold X [0.15] a theme rises if its share of a file's items grows by more than\n X (relative), falls if it shrinks by more than X, else steady. A theme\n that starts at 0 and grows is \"rise\". If either bin holds fewer than\n --min-files files, every theme is \"too few\".\n --min-files N [20]\n --before A:B --after C:D date windows [A, B) and [C, D) by creation date for the before/after\n comparison. Or --split DATE: before = every file created earlier,\n after = every file created on DATE or later. In both modes each file in a\n window is read at creation + --age (snapshot mode too: a window compares\n files by when they started, not calendar periods).\n --kinds K1,K2 [every kind, most files first] file kinds to report separately, in this\n order (also the order of the bootstrap draws)\n --share A:B kind A's share of new files among A and B, per period\n --boot N --seed S [3000, 3] bootstrap resamples and numpy default_rng seed\n --today YYYY-MM-DD [the current date] files younger than --age on this day are left out\n of the bins. Set it to make a run repeatable.\n\nEach sample DIR holds rules.json (extract.py) and labels.json (cluster.py). The first --sample\nwins when a file is in more than one (it is counted once when pooling).\n\n --force-themes apply --themes even when its centroids_id differs from the labels'.\n Without it, analyze refuses: cluster numbers from another fit name\n different clusters, and the themes would silently be wrong.\n\nWrites to OUT (empty bins are null, never 0):\n pooled.json {Q, nfiles, tot, med, pf, files_with_rules, overlap}: all samples pooled; per\n period, files, mean and median items per file, and mean items per file in each\n cluster (pf[i][cluster]). Q is the list of periods, whatever the unit.\n themes.json {q, nfiles, tot, med, from_q, to_q, threshold, groups: [{name, kind, v,\n share_from, share_to}], rest, files_with_items, items, samples: {name:\n {history_files, files, files_with_items, items}}}. share_from/share_to are\n FRACTIONS (0.181 = 18.1%). The input to tufte-charts and watercolor-data-art.\n compare.json per sample and kind: \"periods\" (n, mean, median, pf per period) and \"window\"\n (means, medians, and a bootstrap 95% CI of the difference in means, after minus\n before; \"too_few\": true when a side has fewer than --min-files files).\n share.json per sample (with --share): new files per period by kind, counted by each file's\n first commit, whether it has items or not, leaving out files first committed after\n --today; over --periods if given, else every period. \"share\" is in PERCENT\n (87.2 = 87.2%). Also \"files\" (files with a version) and \"history_files\".\n\"\"\"\nimport argparse\nimport collections\nimport json\nimport os\nimport sys\nfrom datetime import date, timedelta\n\nimport numpy as np\n\nsys.path.insert(0, os.path.dirname(os.path.abspath(__file__)))\nfrom common import period_end, period_of, period_range, period_start # noqa: E402\n\n\ndef load(root, k, a, today):\n R = json.load(open(os.path.join(root, \"rules.json\")))\n life = R[\"life\"]\n labels = json.load(open(os.path.join(root, \"labels.json\")))[\"labels\"]\n if len(labels) != len(life):\n raise SystemExit(f\"{root}: labels.json does not match rules.json\")\n files = R.get(\"files\", {})\n first, kind = {}, {}\n for r in life:\n first[r[\"repo\"]] = min(first.get(r[\"repo\"], \"9999\"), r[\"born\"][:10])\n kind[r[\"repo\"]] = r[\"kind\"]\n if a.created == \"first-version\":\n for f in first:\n key = f[: -len(\".tsv.gz\")] if f.endswith(\".tsv.gz\") else f[: -len(\".tsv\")]\n if key in files:\n first[f] = files[key][\"first\"][:10]\n deleted = {}\n for f in first:\n key = f[: -len(\".tsv.gz\")] if f.endswith(\".tsv.gz\") else f[: -len(\".tsv\")]\n deleted[f] = (files.get(key, {}).get(\"deleted\") or \"9999\")[:10]\n return {\"first\": first, \"kind\": kind, \"life\": life, \"labels\": labels, \"deleted\": deleted,\n \"files\": files, \"nrules\": len(life), \"k\": k, \"age\": a.age, \"today\": today}\n\n\ndef count_at(D, when):\n \"\"\"Items alive per file on a date: when(file) -> 'YYYY-MM-DD' or None (file not counted).\"\"\"\n k = D[\"k\"]\n cnt = {f: 0 for f in D[\"first\"]}\n cl = {f: np.zeros(k) for f in D[\"first\"]}\n at = {f: when(f) for f in D[\"first\"]}\n for r, l in zip(D[\"life\"], D[\"labels\"]):\n t = at[r[\"repo\"]]\n if t and r[\"born\"][:10] <= t and (r[\"died\"] or \"9999\")[:10] > t:\n cnt[r[\"repo\"]] += 1\n cl[r[\"repo\"]][l] += 1\n return cnt, cl\n\n\ndef window(s):\n a, b = s.split(\":\")\n return a, b\n\n\ndef main():\n pre = argparse.ArgumentParser(add_help=False)\n pre.add_argument(\"--config\")\n c, _ = pre.parse_known_args()\n ap = argparse.ArgumentParser(description=__doc__, formatter_class=argparse.RawDescriptionHelpFormatter)\n ap.add_argument(\"--config\")\n ap.add_argument(\"--sample\", action=\"append\", required=True, help=\"NAME=DIR, repeatable\")\n ap.add_argument(\"--themes\")\n ap.add_argument(\"--out\", required=True)\n ap.add_argument(\"--mode\", default=\"cohort\", choices=[\"cohort\", \"snapshot\"])\n ap.add_argument(\"--today\", default=date.today().isoformat())\n ap.add_argument(\"--age\", type=int, default=30)\n ap.add_argument(\"--created\", default=\"first-item\", choices=[\"first-item\", \"first-version\"])\n ap.add_argument(\"--period\", default=\"quarter\", choices=[\"quarter\", \"month\", \"year\"])\n ap.add_argument(\"--periods\")\n ap.add_argument(\"--from\", dest=\"from_\", metavar=\"P\")\n ap.add_argument(\"--to\", metavar=\"P\")\n ap.add_argument(\"--threshold\", type=float, default=0.15)\n ap.add_argument(\"--min-files\", type=int, default=20)\n ap.add_argument(\"--before\")\n ap.add_argument(\"--after\")\n ap.add_argument(\"--split\")\n ap.add_argument(\"--kinds\")\n ap.add_argument(\"--share\")\n ap.add_argument(\"--boot\", type=int, default=3000)\n ap.add_argument(\"--seed\", type=int, default=3)\n ap.add_argument(\"--force-themes\", action=\"store_true\", help=\"apply the themes file even if its centroids_id differs from the labels'\")\n if c.config:\n cfg = json.load(open(c.config))\n cfg = {(\"from_\" if k == \"from\" else k.replace(\"-\", \"_\")): (\",\".join(v) if isinstance(v, list) else v)\n for k, v in cfg.items() if not k.startswith(\"_\")}\n ap.set_defaults(**cfg)\n a = ap.parse_args()\n\n os.makedirs(a.out, exist_ok=True)\n today = date.fromisoformat(a.today)\n samples = [(s.split(\"=\", 1)[0], s.split(\"=\", 1)[1]) for s in a.sample]\n for name, root in samples:\n for f, step in ((\"rules.json\", \"extract.py\"), (\"labels.json\", \"cluster.py fit or assign\")):\n if not os.path.exists(os.path.join(root, f)):\n raise SystemExit(f\"{root}/{f} is missing: run {step} on this sample first\")\n labs = {name: json.load(open(os.path.join(root, \"labels.json\"))) for name, root in samples}\n k_lab = max(l[\"k\"] for l in labs.values())\n ids = {name: l.get(\"centroids_id\") for name, l in labs.items()}\n if len(set(ids.values())) > 1:\n msg = \"samples were labelled against different centroids: \" + \", \".join(f\"{n} {i}\" for n, i in ids.items())\n if not a.force_themes:\n raise SystemExit(msg + \". Assign every sample to the same centroids file (cluster.py assign), or pass --force-themes.\")\n print(\"warning: \" + msg)\n if a.themes:\n cfg = json.load(open(a.themes))\n K = max(k_lab, 1 + max(c for t in cfg[\"themes\"] for c in t[\"clusters\"]))\n theme_list = cfg[\"themes\"]\n want, have = cfg.get(\"centroids_id\"), next(iter(ids.values()))\n if want and have and want != have:\n msg = (f\"{a.themes} names the clusters of centroids {want}, but the labels come from {have}. \"\n \"Cluster numbers from another fit mean different things, so the themes would be wrong\")\n if not a.force_themes:\n raise SystemExit(msg + \". Assign the samples to the centroids the themes file was written for \"\n \"(cluster.py assign --centroids ...), or write a themes file for this fit. \"\n \"--force-themes overrides.\")\n print(\"warning: \" + msg)\n elif not want:\n print(f\"note: {a.themes} has no centroids_id, so it cannot be checked against the labels\")\n else:\n K = k_lab\n theme_list = [{\"name\": f\"cluster {c}\", \"clusters\": [c]} for c in range(K)]\n data = {name: load(root, K, a, today) for name, root in samples}\n per = lambda d: period_of(d, a.period)\n\n # cohort: each file at creation + age; snapshot: each file at the end of every period\n for D in data.values():\n if a.mode == \"cohort\":\n when = lambda f, D=D: (date.fromisoformat(D[\"first\"][f]) + timedelta(days=a.age)).isoformat()\n D[\"cnt\"], D[\"cl\"] = count_at(D, when)\n D[\"ok\"] = {f for f, d in D[\"first\"].items() if date.fromisoformat(d) + timedelta(days=a.age) <= today}\n\n if a.periods:\n Q = a.periods.split(\",\")\n else:\n firsts = [d for D in data.values() for d in D[\"first\"].values()]\n if a.mode == \"snapshot\": # periods that have ended by --today\n p = per(today.isoformat())\n if period_end(p) > today:\n p = per((period_start(p) - timedelta(days=1)).isoformat())\n last = p\n else:\n ok = [d for D in data.values() for f, d in D[\"first\"].items() if f in D[\"ok\"]]\n if not ok:\n raise SystemExit(f\"no file is {a.age} days old by {a.today}\")\n last = per(max(ok))\n Q = period_range(per(min(firsts)), last, a.period)\n\n def bin_files(D, q):\n \"\"\"(files, counts, cluster counts) for one sample in one period.\"\"\"\n if a.mode == \"cohort\":\n fs = [f for f in D[\"first\"] if per(D[\"first\"][f]) == q and f in D[\"ok\"]]\n return fs, D[\"cnt\"], D[\"cl\"]\n end = period_end(q).isoformat()\n fs = [f for f in D[\"first\"] if D[\"first\"][f] <= end and D[\"deleted\"][f] > end]\n cnt, cl = count_at(D, lambda f: end)\n return fs, cnt, cl\n\n # ---- per sample, per kind: periods and the before/after windows\n rng = np.random.default_rng(a.seed)\n\n def ci(x, y):\n bs = [rng.choice(y, len(y)).mean() - rng.choice(x, len(x)).mean() for _ in range(a.boot)]\n return [round(float(v), 1) for v in np.percentile(bs, [2.5, 97.5])]\n\n if a.split:\n b0, b1, a0, a1 = \"0000-00-00\", a.split, a.split, \"9999-99-99\" # open-ended; printed as \"before/from DATE\"\n elif a.before and a.after:\n (b0, b1), (a0, a1) = window(a.before), window(a.after)\n else:\n b0 = b1 = a0 = a1 = None\n if a.kinds:\n KINDS = a.kinds.split(\",\")\n else:\n kc = collections.Counter(D[\"kind\"][f] for D in data.values() for f in D[\"first\"])\n KINDS = sorted(kc, key=lambda x: (-kc[x], x))\n if a.split:\n wb, wa = f\"created before {a.split}\", f\"created {a.split} or later\"\n else:\n wb, wa = f\"{b0}..{b1}\", f\"{a0}..{a1}\"\n compare = {\"Q\": Q, \"today\": a.today, \"mode\": a.mode, \"age_days\": a.age, \"before\": [None, a.split] if a.split else [b0, b1], \"after\": [a.split, None] if a.split else [a0, a1]}\n for name, _ in samples:\n D = data[name]\n res = {\"files_with_rules\": len(D[\"first\"]), \"rules\": D[\"nrules\"]}\n print(f\"== {name}: {len(D['first'])} files with items, {D['nrules']} items\")\n for kd in KINDS:\n rows = []\n for q in Q:\n allf, cnt, cl = bin_files(D, q)\n fs = [f for f in allf if D[\"kind\"][f] == kd]\n n = len(fs)\n rows.append({\"n\": n,\n \"tot\": sum(cnt[f] for f in fs) / n if n else None,\n \"med\": float(np.median([cnt[f] for f in fs])) if n else None,\n \"pf\": (sum((cl[f] for f in fs), np.zeros(K)) / n).tolist() if n else None})\n w = {}\n if b0 is not None:\n if a.mode == \"cohort\":\n cnt = D[\"cnt\"]\n else: # snapshot mode still compares files by creation, each read at creation + age\n cnt, _ = count_at(D, lambda f: (date.fromisoformat(D[\"first\"][f]) + timedelta(days=a.age)).isoformat())\n x = np.array([cnt[f] for f in D[\"first\"] if D[\"kind\"][f] == kd and b0 <= D[\"first\"][f] < b1])\n y = np.array([cnt[f] for f in D[\"first\"] if D[\"kind\"][f] == kd and a0 <= D[\"first\"][f] < a1])\n w = {\"n_before\": len(x), \"n_after\": len(y)}\n if len(x) and len(y):\n w.update(mean_before=float(x.mean()), mean_after=float(y.mean()),\n median_before=float(np.median(x)), median_after=float(np.median(y)),\n diff=round(float(y.mean() - x.mean()), 1), ci95=ci(x, y))\n print(f\" {kd} {wb} -> {wa}: n {len(x)} -> {len(y)}, mean {x.mean():.1f} -> {y.mean():.1f}, \"\n f\"median {np.median(x)} -> {np.median(y)}, diff {w['diff']:+.1f} CI {w['ci95']}\")\n if min(len(x), len(y)) < a.min_files:\n w[\"too_few\"] = True\n print(f\" note: fewer than {a.min_files} files on one side of the {kd} window; too few to tell\")\n res[kd] = {\"periods\": rows, \"window\": w}\n compare[name] = res\n json.dump(compare, open(os.path.join(a.out, \"compare.json\"), \"w\"), indent=1)\n\n # ---- pooled across samples, a file counted once (first sample wins)\n owner = {}\n for name, _ in samples:\n for f in data[name][\"first\"]:\n owner.setdefault(f, name)\n overlap = sum(len(data[n][\"first\"]) for n, _ in samples) - len(owner)\n pooled = {\"Q\": Q, \"nfiles\": [], \"tot\": [], \"med\": [], \"pf\": [], \"files_with_rules\": len(owner), \"overlap\": overlap}\n for q in Q:\n vals, cls = [], []\n for name, _ in samples:\n D = data[name]\n fs, cnt, cl = bin_files(D, q)\n for f in fs:\n if owner[f] == name:\n vals.append(cnt[f])\n cls.append(cl[f])\n n = len(vals)\n pooled[\"nfiles\"].append(n)\n pooled[\"tot\"].append(sum(vals) / n if n else None)\n pooled[\"med\"].append(float(np.median(vals)) if n else None)\n pooled[\"pf\"].append((sum(cls, np.zeros(K)) / n).tolist() if n else None)\n json.dump(pooled, open(os.path.join(a.out, \"pooled.json\"), \"w\"))\n print(f\"== pooled ({a.mode}): {len(owner)} files with items ({overlap} in more than one sample)\")\n for i, q in enumerate(Q):\n if pooled[\"nfiles\"][i]:\n print(f\" {q}: files {pooled['nfiles'][i]:5d} mean {pooled['tot'][i]:5.2f} median {pooled['med'][i]}\")\n else:\n print(f\" {q}: no files\")\n\n # ---- themes\n qf, qt = a.from_ or Q[0], a.to or Q[-1]\n i0, i1 = Q.index(qf), Q.index(qt)\n few = min(pooled[\"nfiles\"][i0], pooled[\"nfiles\"][i1]) < a.min_files\n named, groups = set(), []\n for t in theme_list:\n v = [sum(pf[c] for c in t[\"clusters\"]) if pf is not None else None for pf in pooled[\"pf\"]]\n t0, t1 = pooled[\"tot\"][i0] or 0.0, pooled[\"tot\"][i1] or 0.0\n # an empty compared bin has no share at all: null, never 0\n s0 = v[i0] / t0 if pooled[\"nfiles\"][i0] and t0 else (0.0 if pooled[\"nfiles\"][i0] else None)\n s1 = v[i1] / t1 if pooled[\"nfiles\"][i1] and t1 else (0.0 if pooled[\"nfiles\"][i1] else None)\n if few or s0 is None or s1 is None:\n kind = \"too few\"\n elif s0 == 0:\n kind = \"rise\" if s1 > 0 else \"steady\"\n else:\n change = s1 / s0 - 1\n kind = \"rise\" if change > a.threshold else \"fall\" if change < -a.threshold else \"steady\"\n groups.append({\"name\": t[\"name\"], \"kind\": kind, \"v\": v, \"share_from\": s0, \"share_to\": s1})\n named |= set(t[\"clusters\"])\n rest = [pooled[\"tot\"][i] - sum(g[\"v\"][i] for g in groups) if pooled[\"nfiles\"][i] else None for i in range(len(Q))]\n samples_info = {}\n for name, root in samples:\n R_ = json.load(open(os.path.join(root, \"rules.json\")))\n samples_info[name] = {\"history_files\": R_.get(\"nhistory\", R_[\"nfiles\"]), \"files\": R_[\"nfiles\"],\n \"files_with_items\": len(data[name][\"first\"]), \"items\": data[name][\"nrules\"]}\n themes = {\"q\": Q, \"nfiles\": pooled[\"nfiles\"], \"tot\": pooled[\"tot\"], \"med\": pooled[\"med\"],\n \"files_with_items\": len(owner), \"items\": sum(d[\"nrules\"] for d in data.values()), \"samples\": samples_info,\n \"from_q\": qf, \"to_q\": qt, \"threshold\": a.threshold, \"groups\": groups, \"rest\": rest}\n json.dump(themes, open(os.path.join(a.out, \"themes.json\"), \"w\"), indent=1)\n print(f\"== themes, share of a file's items {qf} -> {qt}\" + (f\" (too few files: fewer than {a.min_files} in a compared bin)\" if few else \"\"))\n for g in groups[::-1]:\n pct = lambda x: \" – \" if x is None else f\"{x * 100:5.1f}%\"\n print(f\" {g['kind']:7} {g['name'][:42]:42} {pct(g['share_from'])} -> {pct(g['share_to'])}\")\n if len(named) < K:\n print(f\" ({K - len(named)} clusters in no theme; their items are in 'rest')\")\n\n # ---- share of kind A among new files of kinds A and B\n if a.share:\n A, B = a.share.split(\":\")\n share = {}\n for name, root in samples:\n files = json.load(open(os.path.join(root, \"rules.json\"))).get(\"files\")\n if files is None:\n raise SystemExit(f\"{root}/rules.json has no 'files'; run extract.py from this skill\")\n fc = {k: (v[\"kind\"], v[\"first\"]) for k, v in files.items() if v[\"first\"][:10] <= a.today}\n qs = Q if a.periods else sorted({per(d) for _, d in fc.values()})\n c = collections.Counter((kd, per(d)) for kd, d in fc.values())\n row = {\"Q\": qs, \"files\": len(fc), \"history_files\": samples_info[name][\"history_files\"]}\n for kd in dict.fromkeys(KINDS + [A, B]):\n row[kd] = [c[(kd, q)] for q in qs]\n row[\"share_of\"], row[\"share_vs\"] = A, B\n row[\"share\"] = [round(100 * c[(A, q)] / (c[(A, q)] + c[(B, q)]), 1) if c[(A, q)] + c[(B, q)] else None for q in qs]\n share[name] = row\n print(f\"== {name}: {A} share of new {A} + {B} files: \" + \", \".join(f\"{q} {s}%\" if s is not None else f\"{q} –\" for q, s in zip(qs, row[\"share\"])))\n json.dump(share, open(os.path.join(a.out, \"share.json\"), \"w\"), indent=1)\n\n\nif __name__ == \"__main__\":\n main()\n","text-rule-trends/requirements.txt","numpy>=1.24\nscikit-learn>=1.3\nfastembed>=0.3\n# optional alternative to fastembed (vectors differ slightly):\n# sentence-transformers>=2.2\n","text-rule-trends/examples/readme-sections/themes.json","{\n \"_note\": \"Names for one k=6 fit of README headings from the five repos in github-file-history/examples/readme-package (collected 2026-10-11). The fit is in centroids.json beside this file; assign new README histories to it to reuse these names.\",\n \"centroids_id\": \"sha256:a53dc00063f14671\",\n \"clusters\": {\n \"0\": \"Usage, example, API\",\n \"1\": \"ky API entries\",\n \"2\": \"How-to sections (configuring CORS)\",\n \"3\": \"mime-types API entries\",\n \"4\": \"chalk API entries\",\n \"5\": \"Install and license\"\n },\n \"themes\": [\n {\n \"name\": \"Install and license\",\n \"clusters\": [\n 5\n ]\n },\n {\n \"name\": \"Usage and examples\",\n \"clusters\": [\n 0\n ]\n },\n {\n \"name\": \"How-to sections\",\n \"clusters\": [\n 2\n ]\n },\n {\n \"name\": \"API reference entries\",\n \"clusters\": [\n 1,\n 3,\n 4\n ]\n }\n ]\n}\n","text-rule-trends/examples/readme-sections/package_deps.py","\"\"\"An extract.py plugin: each dependency named in a package.json is one item.\n\n python3 extract.py SAMPLE --extractor plugin --plugin package_deps.py:items --kind-map '{\"package\\\\.json$\": \"package.json\"}'\n\nThe function gets one version's file body and returns a list of item strings.\nBodies that are not package.json (or not valid JSON) give no items.\n\"\"\"\nimport json\n\n\ndef items(body):\n try:\n pkg = json.loads(body)\n except ValueError:\n return []\n if not isinstance(pkg, dict):\n return []\n out = []\n for field in (\"dependencies\", \"devDependencies\", \"peerDependencies\"):\n for name in (pkg.get(field) or {}):\n out.append(f\"{field} {name}\")\n return out\n","text-rule-trends/examples/readme-sections/config.json","{\n \"_note\": \"README section headings over time, each README read as it stood at the end of every year (snapshot mode). Extract with: --extractor regex --regex '^#{1,3}\\\\s+(.+)' --kind-map '{\\\"(?i)^readme\\\": \\\"README\\\"}'\",\n \"mode\": \"snapshot\",\n \"period\": \"year\",\n \"periods\": [\"2014\", \"2016\", \"2018\", \"2020\", \"2022\", \"2024\", \"2025\"],\n \"from\": \"2016\",\n \"to\": \"2025\",\n \"threshold\": 0.15,\n \"min_files\": 3,\n \"kinds\": [\"README\"],\n \"today\": \"2026-10-11\"\n}\n","text-rule-trends/examples/readme-sections/centroids.json","{\"n_init\": 6, \"seed\": 0, \"id\": \"sha256:a53dc00063f14671\", \"k\": 6, \"dim\": 384,\n \"centroids\": [\n [-0.014929066,-0.029990336,0.029460749,-0.056011185,-0.01168455,-0.011503529,0.032022286,0.029737685,0.05259142,-0.0053320886,-0.007848613,-0.053882387,0.027338315,0.027696284,0.014439185,0.011646536,0.004995344,0.027091084,-0.07525015,-0.007823826,0.06900912,-0.02009877,0.0043570767,-0.036659822,-0.030455042,0.03692167,-0.03116404,-0.029133542,-0.03317994,-0.1149286,0.018300632,-0.019311193,0.0394678,0.016519684,-0.014016878,-0.0021350642,-0.024945302,0.033083387,-0.047402915,0.019413993,0.035122715,0.028337013,-0.056138225,-0.032584157,-0.021995353,-0.02619983,-0.03424485,0.014124299,0.036862176,-0.010424649,-0.020144861,-0.022613171,-0.0036971536,0.009460255,0.012845242,0.071779065,0.075029075,0.03268742,0.02272141,0.0072041913,0.03817138,0.03683632,-0.18559512,0.09757289,0.017435811,0.019542143,-0.0144002205,-0.000500572,-0.0048050517,0.02897311,-0.020388015,0.0032640023,0.010610759,0.105591655,0.019173728,-0.019293813,0.017541401,-0.034710344,-0.00041466084,-0.014889619,-0.020076538,-0.027741905,-0.00760574,0.0018391679,-0.006379782,-0.03313671,-0.012771405,0.0037219166,0.049335264,-0.014375639,-0.035594035,-0.01358631,-0.002022519,0.018481117,-0.03365541,-0.0060580727,0.021295115,-0.004247222,-0.03433696,0.3710286,-0.04447802,0.02404571,0.021591293,-0.02791419,0.025510332,-0.034681994,0.007861127,-0.034985013,-0.030909752,0.030220661,-0.0022340142,-0.009589374,-0.00045725686,-0.024810394,0.024367837,-0.00011489067,0.027188126,0.019969583,0.024293778,-0.018665856,-0.024756357,-0.004181601,-0.008453575,0.009114927,0.018326119,-0.03398667,0.0436868,0.117042586,0.0057355817,0.023925442,0.050168045,0.021685531,-0.036071718,0.0010072823,-0.005423613,-0.016042765,0.029617045,-0.0003199786,0.014482769,0.028759256,-0.028653605,-0.06869765,-0.023340171,-0.10094784,-0.008722269,0.109566376,-0.010766982,0.02884875,-0.012237744,-0.02944736,-0.016130043,0.031904873,0.00031340733,-0.022596665,-0.0029206483,0.01768026,0.035634644,0.000474954,0.013600465,0.02878428,-0.015829353,-0.044356704,-0.053091444,0.079391316,-0.025019055,-0.10512639,-0.0033925099,0.016317438,-0.012279844,-0.014886518,0.012993542,0.00641402,-0.030569622,0.015590574,0.0676942,0.021605764,-0.030948317,0.012467429,0.010750517,0.015745675,0.044147808,-0.043298148,-0.043263838,0.018076226,0.030915434,-0.017052963,-0.013730705,-0.008234593,0.052728612,0.031043813,-0.012349817,0.019898951,0.0019431798,-0.014864338,-0.044971492,-0.018754983,-0.0070334286,-0.020496571,-0.029545149,-0.030255444,0.067424476,-0.004050583,-0.017226875,0.0319014,0.03735836,0.017738909,-0.006505092,-0.018108701,0.041398108,0.022202028,-0.04469029,-0.01380782,0.053317495,0.01788736,-0.031337157,-0.035886697,0.013865478,0.041410647,0.0031367058,0.02399747,-0.010242143,-0.03156655,-0.020835973,-0.3948256,0.013261941,-0.0025762909,-0.022873959,0.0066017183,-0.033216406,0.011306877,-0.0022583015,-0.018886335,0.06326868,0.027088802,-0.049686328,-0.022598723,-0.013064279,-0.016175777,-0.0061061657,-0.0023191597,0.0010098154,0.00088870357,0.009969687,-0.015388122,0.017240146,0.0095473,-0.07038981,0.040111758,-0.010721926,0.2229556,0.06078553,0.008258206,-0.018801888,0.020670323,-0.0014511766,-0.025833983,-0.15147392,0.009907085,0.031124389,-0.021076964,-0.020036016,-0.0030777098,-0.05287701,-0.008252445,0.044229433,-0.0057964665,-0.06280094,0.00977836,-0.014212612,-0.03192698,-0.022778932,-0.042193502,0.025693452,0.0107064545,-0.016958011,0.037025608,0.04939004,0.025203813,-0.032352623,-0.032144196,-0.005872692,-0.010668,0.023259146,-0.003543494,-0.0004538374,0.009026041,-0.05649437,0.02042135,0.008180168,-0.017963616,-0.030677464,0.016386814,-0.018940376,-0.0371581,0.07831759,-0.0064483318,-0.024118856,0.0264362,0.015633818,0.01051688,-0.012780032,-0.019938858,-0.02903547,0.026914537,0.019787041,0.0213883,0.0024136049,0.029331194,0.023714982,0.027350593,-0.015412777,0.051109742,-0.048395082,-0.020811059,0.029805709,-0.010232082,-0.044396557,0.01705185,-0.014569534,-0.36114514,0.031592235,0.005690231,0.06252187,-0.035180535,0.049575906,-0.025148321,0.013215098,-0.06458651,0.022041459,-0.015099584,0.024772363,0.018335804,-0.0022280528,0.010765285,0.05283876,0.08807728,-0.052161902,0.035451617,-0.047756385,0.027079923,0.023746258,0.24896792,-0.022860799,0.014730893,0.027901283,-0.017605362,0.028796174,0.03634044,0.020307127,-0.005022759,-0.009305446,0.09618345,0.0010719767,0.01463146,0.014952969,-0.0019936615,-0.009324705,0.004477831,-0.012534999,-0.045852873,0.022570975,-0.059725616,-0.007815625,0.10199279,-0.01249396,-0.010504422,-0.051643644,-0.018963534,0.014243997,-0.0053315694,-0.025847683,-0.01526413,0.042423274,0.039051767,0.01215547,-0.017038519,-0.0040015695,-0.0027605048,-0.00811075,0.0020190922,-0.027889185,0.017201753,0.024604334,0.013959134],\n [-0.06384458,-0.06311654,0.045370873,-0.03058476,-0.021427972,-0.024652237,-0.01228861,0.0026823948,0.017858604,0.020705476,-0.008341394,-0.032805838,0.02531309,0.0308045,0.025909215,0.019969398,-0.0024495083,-0.0043640602,-0.029164847,0.022667712,0.033292618,0.009204854,-0.0124286525,-0.038321063,-0.032588847,0.015374946,-0.013830087,-0.03653458,-0.06249256,-0.13184369,0.030864427,-0.06705038,-0.018747007,-0.0036145481,0.019476242,-0.021188077,-0.014422821,0.011987513,-0.010022065,-0.0017381771,0.04028962,0.010792736,-0.057403687,-0.05128377,-0.0026572498,-0.034841448,-0.0014986305,-0.0018968633,0.06063777,-0.048191138,-0.023593156,-0.023521833,0.012186043,-0.023332149,0.05810038,0.04070552,0.036616825,0.07995333,0.050478242,-0.003793652,0.008744582,0.07378926,-0.18101329,0.0841368,-0.016084917,0.02445252,-0.0179051,-0.026777564,0.0076525803,0.0027664378,-0.025222171,0.016507534,-0.007835466,0.108962774,0.00793703,0.024139,-0.00325605,-0.0075135888,0.013046067,0.03111451,-0.04423825,-0.058006037,0.018525137,0.025866386,0.006898957,-0.004323081,-0.0058516175,0.020066109,0.068534024,0.0004894479,-0.020533971,-0.03358735,0.03680885,0.030661713,-0.03454437,0.028389584,0.0019001444,-0.008157667,-0.09575336,0.39339817,-0.015788691,0.028136484,-0.029933382,0.0029658063,0.000529583,-0.044146158,-0.0015283125,-0.019407853,0.0050362144,0.00781179,0.030545268,-0.05307384,0.016704466,-0.014396803,-0.039074987,0.07472129,0.016040584,-0.011295105,-0.036240116,0.0073686372,0.010722898,0.013434593,0.036774628,0.026642542,0.02363785,-0.032679282,0.02590291,0.13065955,0.0032358977,0.04465757,0.034323912,0.005231621,-0.06882101,-0.017302167,0.013636227,-0.02432278,0.027206415,-0.0053657633,0.00264876,0.001195392,-0.0026678047,-0.101456344,-0.042286415,-0.0038377987,-0.01954093,0.09593875,-0.016237129,0.015045344,-0.048163656,-0.023401631,-0.02810637,0.021567138,-0.03741783,-0.009006268,0.008500284,0.036314808,0.023740746,-0.008619595,-0.0021497174,-0.018191535,-0.026811816,-0.038045783,-0.056798313,0.069337934,-0.016334668,-0.08077738,-0.0110400785,-0.009767615,-0.01936696,0.02110551,0.019369015,0.01540875,-0.043581206,-0.008619249,0.051507663,-0.008420212,-0.0078017614,0.006111971,0.04096769,0.016542621,-0.017942302,-0.024047365,-0.00095844513,0.02195853,0.04292674,-0.038421627,-0.014590794,-0.04010419,0.045219865,0.024802469,-0.05728715,-0.03923957,-0.012381767,-0.06702836,-0.022768939,-0.07161474,-0.019421907,0.0048180083,0.06154111,-0.023806885,0.13184987,0.048819028,-0.029223677,0.032623578,0.023218058,0.024741698,-0.0041219033,-0.017243735,0.023196388,-0.017869491,-0.07551498,-0.029311398,0.072385706,0.0019595919,-0.05209834,-0.002301247,0.008278141,0.069906704,0.035110377,0.0007566321,-0.020275783,-0.004091571,-0.023017416,-0.36002183,0.037827518,0.011621494,-0.035290163,-0.042988613,-0.030136207,-0.00058186596,-0.0077153896,-0.016842315,0.05871538,0.0813856,-0.009144929,-0.0065584444,-0.014961174,0.03688362,-0.035992615,-0.02475768,-0.002801722,-0.03797689,0.020112693,-0.014623594,-0.024101723,0.000043884967,-0.093709506,0.0076781926,-0.0034511378,0.19684085,0.025311695,0.043849263,-0.03705388,0.0475339,0.0850939,-0.010453292,-0.10805921,0.018371826,0.0316794,-0.008302409,0.05725121,0.0075366013,-0.03697945,-0.026109444,0.0047018905,-0.008106095,-0.06700678,0.048535854,-0.0219405,-0.02109671,-0.0121398615,-0.0077988952,0.0073113167,0.018747244,0.009866721,0.04481594,0.030036284,0.054936964,-0.023912612,-0.011221127,-0.0056308974,0.055901084,0.02665899,0.030416168,0.00059216586,-0.0021934088,-0.040424272,0.035876546,-0.006444462,0.010603672,-0.008651144,0.03172571,-0.011734,-0.025558555,0.04224081,-0.0007900055,0.04793144,0.0027446854,0.049129866,-0.015885754,-0.011125587,0.004689618,-0.013665117,0.0123568475,-0.032303564,0.04169946,0.01171258,0.031724393,0.03614896,-0.011783978,0.033457033,0.041943528,-0.017451847,-0.0376137,-0.0043885508,-0.05124452,0.044470716,0.05540329,-0.03680506,-0.31606808,0.010660491,-0.0027697552,-0.0030328443,0.00040438728,0.05325639,0.00047898613,-0.030148288,-0.06764634,-0.0077267243,-0.03918673,-0.0020889484,-0.026612518,-0.017192092,0.026685134,0.045893885,0.03049597,-0.019360479,0.02526997,-0.01894572,0.0653783,0.044150427,0.18110393,-0.029868256,0.04052856,0.03352334,-0.0019638124,0.032263022,0.08460867,-0.036882803,-0.024217682,0.012888264,0.071506694,-0.028132439,0.034150742,-0.013494748,-0.022054112,-0.03301114,0.00816367,-0.06262155,-0.014346388,0.041795075,0.012328666,0.0036147498,0.06767705,-0.004329416,-0.018191162,-0.06846224,-0.0021717008,0.024330603,-0.029610654,-0.03068562,-0.018265149,0.023213984,0.05449552,0.03946392,0.00990373,-0.00644825,-0.0253314,-0.009702412,0.009248786,-0.062357046,0.05967453,-0.019317029,-0.013003971],\n [-0.04591561,-0.013622777,0.015199327,-0.028069872,-0.030086013,-0.026294995,-0.046886638,-0.02572163,-0.008199092,-0.026679281,-0.0034648178,0.0068755792,0.02987921,0.029953726,0.048327252,0.015876835,0.011238486,0.050116204,-0.011722071,0.019055113,0.021566622,0.02012381,0.0070072617,-0.031027265,-0.09668778,0.046321195,-0.018281473,-0.026921872,-0.055759616,-0.14666848,0.02142966,-0.08189542,-0.014125471,0.01402724,0.010109915,-0.03556183,0.01716712,-0.007050329,-0.013855649,0.027561933,-0.00022982161,0.005015807,-0.02094068,-0.018404918,-0.047252327,-0.049442936,0.022035081,0.013737356,0.009168473,-0.031691335,0.017877469,0.016270628,0.0012279017,-0.0065397983,0.013042343,0.07665052,0.0321386,0.07546317,0.04150721,0.004456831,0.008887225,0.0015964126,-0.20723562,0.097247355,0.004667932,0.051223155,0.004688963,-0.020777078,0.026065467,0.0033831664,-0.0584375,0.030910322,-0.00955992,0.049185246,0.047703624,0.025969356,-0.013283116,0.010156199,0.04656741,-0.020714814,-0.028793363,-0.019597782,-0.04041635,0.0098772915,-0.003789098,-0.043393046,-0.03292845,-0.0039145164,-0.009825406,-0.0011103343,-0.039297167,-0.10360528,0.047439735,0.036130205,-0.058286708,0.015985578,-0.0030919472,-0.004679931,-0.07762823,0.36027646,-0.001401763,0.025931042,0.013238015,-0.02942141,0.030200293,-0.042757444,-0.025199804,-0.00091972697,0.033573814,0.021811726,-0.010866095,-0.0018300272,0.0068095075,-0.03373855,-0.0061988113,0.0038202505,-0.043595396,0.0030118437,0.03188572,0.0035838464,0.016157717,-0.013450971,0.021417562,0.039211977,0.018272536,-0.009991841,0.033734806,0.0885346,-0.0035638937,0.08961823,0.017076196,0.011796928,-0.034675032,0.023517024,0.020370763,0.008810913,0.00092638686,0.008892232,0.030339709,0.03905437,-0.024243666,-0.055599887,0.0068975175,-0.030967202,0.018721819,0.06371429,-0.053601306,0.08327933,-0.048224736,-0.053258535,-0.012356773,0.027392246,-0.054823246,-0.019340869,-0.05866813,0.0332599,0.018498816,-0.021600146,-0.005494866,-0.0124596385,-0.0034089962,-0.024194293,0.0030477848,0.098200314,-0.009973779,-0.11680489,0.014081789,0.0076227067,-0.0005614432,0.0077416655,-0.05131532,0.0055030817,-0.0028767544,-0.023234563,0.07528068,0.010684647,-0.070231095,0.03951265,-0.024710206,-0.0037586747,-0.006769695,-0.04143101,-0.0061054137,0.028703686,0.0027076942,-0.026078392,-0.06674858,-0.07744663,-0.011295501,-0.05700821,-0.07633437,-0.02093833,-0.06352058,-0.07180138,-0.0067923744,0.00030199927,-0.019960279,-0.005416276,-0.006864253,-0.010900847,0.12947534,0.02176944,0.0037702147,0.04779771,0.0024997268,0.013827231,0.007752383,-0.0222826,0.029756455,-0.03805842,-0.037752964,0.006516122,0.10282794,0.010816054,-0.04886258,-0.012841511,0.016226819,0.0665477,0.024281433,0.050469227,0.014188685,0.038872283,-0.045544136,-0.36007178,0.033053093,0.0072286497,-0.015720794,-0.007580418,-0.03530704,0.015110791,0.0034182442,-0.024446564,0.05490468,0.0982538,0.014678184,0.0028911424,0.023550706,-0.013106378,0.027112331,-0.0028082794,-0.015522709,-0.026192063,0.019245403,-0.0045775794,-0.046632133,0.03609297,-0.054034133,0.011284909,0.0045898813,0.17975992,0.0099659115,0.030901026,-0.032746706,0.065242045,0.024397394,-0.02355959,-0.13210262,-0.018796628,0.055905845,0.017022293,0.048898302,0.06919207,-0.008726153,0.00024212162,0.04146914,0.016438011,-0.06482682,0.017771132,-0.03730464,-0.018754963,0.01715255,-0.04425646,-0.0040486325,-0.015502964,0.014124217,0.011221409,0.022737771,0.019235399,-0.010330603,0.01678962,-0.007333039,0.024885774,0.021541724,-0.0017188317,-0.015405086,-0.022312287,-0.052950248,0.069394395,-0.008925862,-0.006370316,-0.030676436,0.048182752,-0.038408432,-0.045835804,0.019839548,-0.0022648335,0.07449934,-0.0059301937,-0.01899253,-0.008103183,-0.017096449,-0.005112106,-0.05071714,0.03277574,-0.020108225,0.0075965645,0.007333112,0.04355635,0.011529798,0.008303893,0.012851196,0.0111759,0.015297858,-0.03248028,0.0019366241,-0.0030563525,-0.020290168,0.0588074,0.02697689,-0.30586976,0.0015030464,-0.0016671248,-0.012002794,0.0023220754,0.010250779,0.038222075,-0.008303914,-0.10975587,0.059595693,-0.048981044,0.031961963,-0.042358633,-0.0012974764,0.037287228,0.073350005,0.014582289,-0.014051806,-0.015809827,-0.080840796,0.05514844,0.02083245,0.18660231,-0.01791281,-0.0053889942,0.08394452,0.019910783,0.062278975,0.036082037,0.034421403,0.017501758,-0.014342832,0.045394264,-0.005549963,0.019189043,0.044966698,0.0074347868,-0.016862081,0.050524138,0.0010478643,-0.034384917,0.022559708,-0.012159464,-0.0036979339,0.058979146,-0.03997935,-0.018753946,-0.020896375,0.02130583,0.02486513,-0.047663547,-0.05939926,0.010851425,0.015296695,0.020328172,0.02870203,-0.0054777944,-0.028776485,-0.05092959,-0.00048598158,0.04634749,-0.023368306,0.064892694,0.021902032,0.006420518],\n [-0.0761108,0.0028936442,0.040883653,-0.021315144,0.016677909,-0.0063794656,-0.021437459,-0.0004045487,0.013458765,-0.03341763,0.0042565316,-0.04304974,0.008371787,0.0064723548,0.100946784,0.009092308,0.0304951,0.025754843,-0.08675882,0.0046449043,0.060764775,0.0009040764,-0.008506996,0.010157532,-0.015387141,0.022347692,0.017551443,-0.036306605,-0.04654845,-0.16014193,0.002937365,0.04822043,0.022589326,0.030336337,-0.011560625,-0.019520095,-0.07016791,0.009648911,-0.017149312,0.032550134,0.03307959,-0.023725213,-0.01765551,-0.040081162,-0.02951175,-0.089433774,-0.0111675905,-0.003084952,-0.019308735,-0.076551,-0.035211526,0.017796103,0.019155245,-0.017777443,0.006578838,0.0822726,0.013135783,0.03819775,-0.013093385,-0.018873917,0.021391243,0.0006942665,-0.14946558,0.08314663,-0.013867921,0.05564831,-0.04581101,-0.004243433,0.017125865,-0.0078055062,-0.029954227,0.036009394,-0.07432401,0.061412007,0.017127614,0.036538444,-0.00604513,-0.03442382,0.014801282,0.0084901005,-0.048566196,-0.089337684,0.024594385,0.0019556775,0.046542455,-0.019223955,-0.006809172,0.016255824,0.025141487,-0.0073581114,-0.077926844,-0.037839234,0.03357344,0.053763133,-0.012181429,0.0081605995,0.029795544,0.026271323,-0.0119176265,0.38056946,-0.031896427,0.0049580135,-0.001477999,-0.03474603,0.019093176,-0.006238138,-0.0041145235,-0.03599303,-0.0018393667,-0.04849274,-0.03827298,-0.023564838,-0.010129843,-0.0012863948,-0.03896594,0.0098636765,-0.032352217,0.032048605,0.010648427,0.001722324,-0.011685628,-0.023329822,0.02008326,0.0028840199,0.023969684,0.0042928876,0.020936875,0.043753285,0.009947101,0.05247941,0.011245624,0.06751315,-0.05769978,-0.025106221,0.015195069,0.017908849,-0.021754205,-0.063094676,0.022973683,-0.022921808,-0.03800794,-0.07838976,0.010300526,-0.04161201,-0.07881091,0.10636375,-0.06474299,0.03878157,-0.029042374,0.023257567,-0.022031587,0.059640087,-0.060317967,-0.01799155,0.019483456,0.011466353,0.050674066,-0.014662741,-0.045731444,-0.030383445,-0.003597804,-0.00484406,-0.04309465,0.09063093,-0.014014983,-0.08335959,0.0054688742,0.04022354,0.011235903,-0.016834505,0.007826894,0.000036471603,-0.003917243,-0.00066690386,0.038108196,-0.030840738,-0.048836496,-0.027899817,0.021671986,0.009106692,-0.015128911,-0.0031009235,-0.022991458,-0.0032913636,-0.0037454874,-0.015899613,0.01659185,-0.018365402,0.062207397,0.019375699,0.014392204,0.02698692,0.0075042336,-0.053970568,-0.02382137,-0.016675817,0.0020544573,0.011166593,0.03599058,-0.041535426,0.091686055,0.04273533,0.0033178795,-0.0132266525,-0.03162826,-0.0036414077,-0.022923619,-0.01950068,0.037783507,-0.018844904,-0.06329864,-0.0031601354,0.08264194,0.022590866,-0.046855815,0.0032911017,0.057485417,0.054444626,0.031672787,0.03696889,-0.053160347,-0.05786446,-0.08957865,-0.34738967,-0.022134384,0.015893875,-0.06078893,-0.07580337,-0.047570225,0.023034897,-0.018201724,-0.02899779,0.03858805,0.044104222,-0.059969135,-0.040973883,-0.025129594,0.026713612,0.009742187,0.020892786,0.0006343261,-0.016424274,0.015715472,0.0021353588,-0.010030944,0.010529224,-0.0041657723,-0.0069715152,-0.0053466894,0.17308664,0.023622436,0.0031249223,-0.022827014,0.055439185,0.038676754,0.018083137,-0.07449337,-0.024229169,0.046101388,-0.022734718,0.034994353,0.047274843,0.059799228,0.015541922,0.015826963,0.04491754,-0.06504801,0.05088241,-0.07911593,0.027193556,-0.04341097,-0.039234087,0.045503773,-0.034740437,0.0064909775,0.043188233,0.03232538,-0.005573389,-0.0035837216,-0.05235915,-0.04204614,0.013704204,-0.022724414,0.031756654,0.006299093,0.030574849,-0.03971879,0.005298837,-0.006446062,-0.0022956987,-0.055199344,0.020016223,-0.03555969,0.010459958,0.06365522,0.025713239,0.006018867,-0.03316899,0.030006133,-0.044885837,-0.023456847,-0.009673857,-0.0023102153,0.010408721,0.04188217,0.045905817,0.06713489,0.0032605622,0.011735782,0.0027471934,-0.01249357,0.025111694,0.017101677,-0.0077630575,0.005731139,-0.013265763,0.03192239,0.01606162,-0.02435171,-0.30298766,0.06648487,0.059874207,0.0052442206,-0.006447801,0.021053204,-0.0052225594,-0.08608316,-0.08299895,0.025808038,0.038179506,0.028105693,0.0493452,-0.018328689,0.02779174,0.03900614,0.058171704,-0.013830639,0.058375545,-0.055917863,0.028941948,0.009053211,0.21499,0.002026781,0.019022789,0.047440466,0.0073044323,0.02647888,0.050882332,0.057247963,0.035322025,-0.030746391,0.12133159,-0.048606522,0.02082377,0.0649314,0.032697663,-0.006165564,0.022758953,-0.032739323,-0.0020011351,0.038369585,-0.070319,0.023308886,0.022906398,-0.026837992,0.0014024321,0.005037522,-0.020437721,0.016377272,-0.05846992,-0.010312191,-0.041506704,-0.0055138394,0.021882134,0.013323116,-0.0058367117,0.009102986,-0.018949857,0.04326336,0.047074225,-0.019518323,0.029734235,0.049823634,0.03350764],\n [-0.05795525,-0.03358782,0.0085270805,-0.024324734,-0.008455378,0.031830292,-0.018475927,0.017634336,0.029499002,0.0020548203,-0.028310653,-0.03587227,0.0042667394,0.022627095,0.013962917,0.049144078,0.024832409,0.055934615,-0.06559923,-0.022548271,0.073548324,0.0035418533,0.018395688,-0.09554905,-0.014017099,0.02246867,-0.040957347,-0.026482522,-0.048558652,-0.1828904,-0.037755307,-0.030869784,0.0654471,-0.030355409,0.030651907,-0.004605883,-0.028027497,0.037916146,-0.0312043,-0.0077545783,0.010758545,-0.01817423,-0.057814617,-0.001112346,-0.020530008,-0.041035883,-0.0061530247,-0.008452154,0.034300584,-0.0222059,0.017117353,-0.03555884,-0.021718813,0.005992003,-0.007129607,0.056429148,0.082502656,0.03736738,0.016593313,0.0000012693124,0.0196267,0.037367627,-0.16566658,0.1085466,0.028793933,0.00911296,-0.020682536,-0.0329301,0.011480351,0.033363845,-0.016745154,0.032133408,0.027392544,0.064946145,0.013471918,-0.014143577,-0.0002307745,-0.060581323,-0.02928931,-0.0650653,-0.02250934,-0.0038920029,0.005544636,0.0064188023,0.037651412,0.010693218,-0.04395778,0.0013206003,0.05147291,-0.00949982,0.0075283777,-0.027739001,0.021008225,0.008551215,-0.08247321,-0.01630335,-0.009692986,-0.011518339,-0.00487499,0.44012266,-0.038195934,0.0029633923,0.05167082,-0.020105114,0.048641056,-0.024714692,0.02915363,-0.03320974,-0.051973633,-0.017374719,0.008605014,-0.034022,0.057983845,-0.034066267,-0.021190364,-0.012908996,-0.0024291824,-0.005957116,0.011868527,0.01853459,-0.0035573596,0.0020152037,-0.0045159874,-0.01868029,0.018951142,-0.0019077179,0.02293359,0.12959635,0.011737791,0.046310708,0.0056511094,0.0209064,0.005588882,0.015764032,-0.03036513,0.0067697302,0.028655535,0.009046772,0.019658446,0.014019378,-0.009415393,-0.015463241,0.02697709,-0.030511249,0.05719911,0.035751265,-0.061021607,0.024205023,-0.02691817,0.008601741,-0.059764635,0.049206674,-0.0056289155,-0.0052237324,-0.047675293,0.03416186,0.033678196,-0.02536861,-0.014352486,0.040171903,0.008479956,-0.061854493,-0.012661679,0.041456807,-0.009742462,-0.08066405,-0.017902464,-0.027941007,-0.017087657,-0.041489057,-0.010427782,0.024783494,0.017877398,-0.025196014,0.07373539,0.014590045,-0.037794992,0.0053318026,0.040010877,0.018325074,0.027080866,-0.022065153,-0.017301181,0.015241997,0.026890874,-0.043394953,0.029975558,-0.0011953164,0.050590687,0.056658044,-0.020540606,-0.028310291,-0.040579952,-0.019269751,-0.02756139,-0.01610637,-0.019636111,-0.049944073,0.005751505,-0.021178404,0.074137926,0.014112146,-0.021799222,0.014453987,0.027433071,0.009003325,0.01596147,-0.046801705,0.03598464,-0.0003064607,-0.057855092,0.009396777,0.07059073,-0.011738123,0.009854348,-0.010224822,0.012247399,0.005804057,0.011740073,0.0042415876,-0.023429405,-0.032163464,-0.09663451,-0.381236,0.042734474,0.037665058,-0.022422852,0.0001853543,-0.025108626,0.012335595,-0.009589268,0.025037602,-0.00064611575,0.061105113,-0.008405903,0.02047366,-0.035347123,0.01847343,-0.0039018632,0.008593756,0.022161303,0.040643085,0.0024173677,-0.021329831,0.0109886415,0.012261113,-0.05462268,0.05798259,-0.013637014,0.18692109,0.027310275,0.0030312445,-0.030214705,0.011163857,0.046969883,-0.035265915,-0.07317142,0.040488657,0.0455028,-0.030627666,0.026877536,-0.04258353,-0.010140798,0.03138053,-0.001329469,-0.0016153674,-0.06634067,-0.004206679,-0.02490645,0.009087077,-0.023290249,-0.027847197,0.03288979,0.004627329,-0.015369142,0.025635535,0.0646565,0.0056636673,0.0007490995,-0.0047779675,0.014785841,-0.01814097,0.0119097205,-0.0041857,-0.050686814,0.0016465841,-0.043537006,0.0062912027,0.038509253,0.0031826117,-0.050781656,0.039065797,0.0027451986,-0.015598093,0.0432132,-0.00678622,0.012656378,-0.00240849,0.040516816,0.023866445,0.03704546,-0.00016547268,-0.02735129,0.011987658,0.04101918,0.034318913,0.0042893463,0.019425064,0.015015337,0.022792045,-0.02513469,0.051778555,0.00731026,-0.032818854,0.02041857,0.020849997,-0.015898168,0.030602207,-0.019548858,-0.33063132,0.016059218,0.007364223,0.019246977,-0.009423509,0.033399988,0.035705846,-0.020348722,-0.057756457,0.009467951,-0.04782446,0.0034967815,0.0035499642,-0.039038062,0.0061237332,-0.01183495,0.09403299,-0.035895642,0.045884255,-0.059037887,0.007178352,0.0030413293,0.20488884,-0.022615457,-0.02645721,0.01954027,0.039080214,0.010774016,0.05164415,0.030353947,0.008294914,0.02874016,0.0759374,-0.011787009,-0.014756346,0.075253755,-0.014812266,-0.030342788,0.0034751687,-0.020761432,0.015172087,0.015294199,-0.07371124,-0.011661654,0.049235977,-0.019970141,-0.0038328,-0.014331646,0.007591291,-0.017636556,-0.030043472,-0.028138913,-0.022644961,0.01630295,0.01538737,0.008764756,-0.05195763,0.0023051526,0.0059917956,-0.032057658,0.00054504274,-0.044502202,0.023839485,0.0055040587,0.004991771],\n [-0.01865073,-0.034801643,0.03252173,-0.062015407,-0.0035251947,0.033312935,-0.0107205445,0.04298088,0.019526485,0.002206602,-0.008961233,-0.062123638,0.028646898,0.008275763,0.009907156,0.015510203,0.026227674,0.05436744,-0.040344093,-0.00022375934,0.029352404,-0.03248271,-0.030756414,-0.04971878,-0.04563274,0.041611034,-0.0333881,-0.016810957,-0.08249218,-0.096958846,-0.0067567197,-0.025579564,0.032511294,0.014154168,0.028624866,-0.028062051,-0.024873853,0.013572092,0.00025667727,-0.0050407858,0.032958895,0.008564729,-0.07372884,-0.02538887,0.017572694,-0.025189858,-0.008980105,-0.0008760654,0.065945916,-0.028995933,-0.016735364,-0.0011448058,-0.012699382,0.019589402,-0.025637113,0.041278616,0.062446903,0.03115662,0.026881943,0.049949177,-0.0004314892,0.033508807,-0.24479286,0.123229936,0.02549709,0.032711416,0.00060117926,-0.040495608,-0.00085491163,0.016338378,-0.036291953,0.013431391,0.00007013743,0.06517368,0.030754536,0.0007408633,0.016157614,-0.03964767,0.0014455911,-0.012749245,-0.01608061,-0.027358381,-0.002735838,0.016668702,-0.03133399,-0.027447384,-0.029033575,0.029962648,0.05078071,0.014433332,-0.017591937,-0.028581453,0.013705296,0.01283677,-0.015415344,-0.01690432,-0.010929304,-0.018029127,-0.052696522,0.35236624,-0.02606324,0.009323284,-0.010151878,-0.028769555,0.036431402,-0.021825531,-0.018647961,0.004615602,-0.047007337,-0.0021367199,0.0013391883,0.003873461,0.018011594,-0.010394036,0.02765797,0.041746307,0.005595307,0.012956117,0.030789262,-0.025183512,-0.041593626,0.010378885,0.0046351617,0.02002652,0.025499305,-0.0708015,0.011599919,0.12838766,0.01397853,0.08347593,0.040856253,0.025388045,-0.039827015,0.032751426,0.017867887,0.0069700265,-0.016727785,0.008783782,-0.009448293,0.016509812,-0.031020708,-0.05468115,0.00964775,-0.078156404,-0.0045936285,0.05014071,-0.009843292,0.044862173,-0.042519003,-0.050690748,-0.0060641323,0.043931358,-0.025434142,-0.006800883,0.021956695,0.014411794,0.060338642,-0.01270659,-0.055077817,0.011471849,0.0011690855,-0.041902967,-0.030516531,0.07932898,-0.028695395,-0.10538739,0.005958735,0.0019832468,-0.0010253405,-0.0033903976,0.024172781,0.015741289,-0.038927585,0.028262746,0.06763995,-0.0022767237,-0.0124814,0.025589619,0.005647749,-0.0016495793,-0.015857536,-0.019864693,-0.050885,-0.008632444,0.035741445,-0.021607142,-0.044910308,-0.015308075,-0.014928573,0.029496541,-0.027434867,0.063973226,0.020702124,-0.03567295,-0.02459513,-0.036094483,-0.025934294,-0.029416172,-0.012313842,-0.046260666,0.06400321,0.028941374,-0.025831614,0.023255624,0.030678507,0.010729928,-0.014496871,-0.014943302,0.01474688,0.015755394,-0.02764531,0.025108192,0.09495977,0.00723613,-0.031453326,-0.01757865,0.009004831,0.04761319,0.025581364,0.014717121,-0.016597234,-0.0016248431,-0.021694917,-0.3891123,0.044944294,-0.019644659,0.0053223977,0.015825992,-0.034474846,0.032974828,-0.013341782,-0.022257155,0.052529544,0.10393891,-0.0526535,-0.017445123,0.047109094,-0.01689107,0.024868114,0.02849796,0.0063774697,0.022725958,0.007786293,0.005777883,0.028393803,0.0034983736,-0.0520528,0.06458126,-0.01246301,0.2105326,0.025831448,0.034867544,0.003868707,0.03917323,0.0046688733,0.0033564232,-0.17293726,-0.0034772616,0.04979137,-0.037244603,0.015736237,-0.024972918,-0.0456472,-0.0063459035,0.051178116,-0.012899738,-0.0632235,-0.007388036,-0.030973949,-0.015376164,0.030156307,-0.04576868,-0.016199095,0.0097097745,-0.0032549058,0.053443942,0.04991388,-0.00465086,-0.033886332,-0.025268873,0.015325752,0.00046300728,0.0111954305,0.02230397,-0.009309138,0.010398429,-0.05556526,0.027003672,0.004861186,-0.021297915,-0.013079177,0.071534194,-0.053257834,-0.041877158,0.01409442,-0.0038968688,0.0036484161,0.021684214,0.040171713,0.0154017545,-0.0057376376,-0.0061548604,-0.032288644,-0.0006424081,-0.023775226,0.033289786,0.0015034858,0.042862844,0.029716248,0.022150798,-0.0250202,0.0067621144,0.014897308,0.0073861475,0.023870097,-0.02025393,-0.047137357,0.008919014,-0.024171557,-0.34867263,0.0072663524,0.008492188,0.026257595,-0.04052178,0.016292809,0.013018189,0.027651321,-0.03961438,0.0037912305,-0.014144121,0.029756311,-0.018650401,-0.0224152,0.021628492,0.032773048,0.11797445,-0.038522467,0.015974158,-0.054831382,0.0070461114,0.031834435,0.19146349,-0.024061969,0.030801538,0.015726075,-0.008063905,0.03448311,0.04437003,0.00011346124,0.0219925,0.015573472,0.08909263,-0.01812093,0.00018476737,0.053240072,-0.04539701,0.01705896,-0.0010077829,-0.0062543913,-0.04072893,-0.0063798935,-0.019108143,0.0043962053,0.07105705,-0.05215003,-0.021669554,-0.03897759,-0.046853025,0.023297459,0.003797815,-0.0058032316,-0.009330756,0.022412634,0.014677532,0.004263949,-0.0134416865,-0.01405891,-0.0149922315,-0.006447708,-0.00019579126,-0.04832036,0.032946046,0.052220374,-0.012778989]\n ]\n}\n","text-rule-trends/examples/made-up/vers/example_project-l__CONTRIBUTING.md.tsv","2134bd5850695a45d8475f4ca3e561c5c073de0d\t2026-09-28T10:00:00+00:00\t0\tIyBDb250cmlidXRpbmcKClRoYW5rcyBmb3IgaGVscGluZy4KCiMjIFJ1bGVzCgotIEFsd2F5cyBydW4gdGhlIHRlc3RzIGJlZm9yZSB5b3Ugb3BlbiBhIHB1bGwgcmVxdWVzdC4KLSBOZXZlciBza2lwIGEgZmFpbGluZyB0ZXN0IHdpdGhvdXQgYW4gaXNzdWUgbnVtYmVyLgotIFByZWZlciBzbWFsbCBmdW5jdGlvbnMgb3ZlciBsb25nIG9uZXMgd2l0aCBtYW55IGJyYW5jaGVzLgotIERvY3VtZW50IGV2ZXJ5IG5ldyBvcHRpb24gaW4gdGhlIHJlZmVyZW5jZSBwYWdlLgotIEtlZXAgdGhlIFJFQURNRSBzaG9ydCBhbmQgbGluayB0byB0aGUgZG9jcyBmb3IgZGV0YWlsLgotIERvIG5vdCBjb21taXQgZ2VuZXJhdGVkIGZpbGVzIG9yIGxvY2FsIHNlY3JldHMuCi0gV3JpdGUgY29tbWl0IG1lc3NhZ2VzIGluIHRoZSBpbXBlcmF0aXZlIG1vb2QuCi0gTmV2ZXIgcHVzaCBkaXJlY3RseSB0byB0aGUgbWFpbiBicmFuY2guCg==\n4ad0ce774b3984cf752e136395eaf3b476af0299\t2026-08-27T10:00:00+00:00\t0\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\nc57f7b2c4c8bc21ef0837e27e4225c92adf346b5\t2026-07-26T10:00:00+00:00\t0\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\n1faa9b86bfdb49d203a1fb8b6f65421fb603c6c6\t2026-06-25T10:00:00+00:00\t0\tIyBDb250cmlidXRpbmcKClRoYW5rcyBmb3IgaGVscGluZy4KCiMjIFJ1bGVzCgotIEFsd2F5cyBydW4gdGhlIHRlc3RzIGJlZm9yZSB5b3Ugb3BlbiBhIHB1bGwgcmVxdWVzdC4KLSBOZXZlciBza2lwIGEgZmFpbGluZyB0ZXN0IHdpdGhvdXQgYW4gaXNzdWUgbnVtYmVyLgotIFByZWZlciBzbWFsbCBmdW5jdGlvbnMgb3ZlciBsb25nIG9uZXMgd2l0aCBtYW55IGJyYW5jaGVzLgotIERvY3VtZW50IGV2ZXJ5IG5ldyBvcHRpb24gaW4gdGhlIHJlZmVyZW5jZSBwYWdlLgotIEtlZXAgdGhlIFJFQURNRSBzaG9ydCBhbmQgbGluayB0byB0aGUgZG9jcyBmb3IgZGV0YWlsLgotIERvIG5vdCBjb21taXQgZ2VuZXJhdGVkIGZpbGVzIG9yIGxvY2FsIHNlY3JldHMuCi0gV3JpdGUgY29tbWl0IG1lc3NhZ2VzIGluIHRoZSBpbXBlcmF0aXZlIG1vb2QuCi0gTmV2ZXIgcHVzaCBkaXJlY3RseSB0byB0aGUgbWFpbiBicmFuY2guCg==\n","text-rule-trends/examples/made-up/vers/example_project-k__CONTRIBUTING.md.tsv","6f26428f7f1658f169522f80197edc38ba0cd508\t2026-08-21T10:00:00+00:00\t0\tIyBDb250cmlidXRpbmcKClRoYW5rcyBmb3IgaGVscGluZy4KCiMjIFJ1bGVzCgotIEFkZCBhIHRlc3QgZm9yIGV2ZXJ5IGJ1ZyB5b3UgZml4LgotIE5ldmVyIHNraXAgYSBmYWlsaW5nIHRlc3Qgd2l0aG91dCBhbiBpc3N1ZSBudW1iZXIuCi0gVXNlIGZvdXIgc3BhY2VzIGZvciBpbmRlbnRhdGlvbiBpbiBldmVyeSBmaWxlLgotIFVwZGF0ZSB0aGUgY2hhbmdlbG9nIHdoZW4geW91IGNoYW5nZSBiZWhhdmlvdXIuCi0gRG9jdW1lbnQgZXZlcnkgbmV3IG9wdGlvbiBpbiB0aGUgcmVmZXJlbmNlIHBhZ2UuCi0gTmV2ZXIgcHVzaCBkaXJlY3RseSB0byB0aGUgbWFpbiBicmFuY2guCg==\nd4658e11b173f211929bcbf2f3537d1a0016fd79\t2026-07-20T10:00:00+00:00\t0\tIyBDb250cmlidXRpbmcKClRoYW5rcyBmb3IgaGVscGluZy4KCiMjIFJ1bGVzCgotIEFkZCBhIHRlc3QgZm9yIGV2ZXJ5IGJ1ZyB5b3UgZml4LgotIE5ldmVyIHNraXAgYSBmYWlsaW5nIHRlc3Qgd2l0aG91dCBhbiBpc3N1ZSBudW1iZXIuCi0gVXNlIGZvdXIgc3BhY2VzIGZvciBpbmRlbnRhdGlvbiBpbiBldmVyeSBmaWxlLgotIFVwZGF0ZSB0aGUgY2hhbmdlbG9nIHdoZW4geW91IGNoYW5nZSBiZWhhdmlvdXIuCi0gRG9jdW1lbnQgZXZlcnkgbmV3IG9wdGlvbiBpbiB0aGUgcmVmZXJlbmNlIHBhZ2UuCi0gTmV2ZXIgcHVzaCBkaXJlY3RseSB0byB0aGUgbWFpbiBicmFuY2guCg==\n32ebc0da645471947136eee236a4eb16e9244a3d\t2026-06-19T10:00:00+00:00\t0\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\n3b89376cda8add395c116c31a0d76b78fd9e02e2\t2026-05-18T10:00:00+00:00\t0\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\n","text-rule-trends/examples/made-up/vers/example_project-j__CONTRIBUTING.md.tsv","b998af69319fe4c6f09f51f921ca83f554f75765\t2026-06-11T10:00:00+00:00\t0\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\nc5d062205b213c6b4258e677c2e6e9843fa4ec0d\t2026-05-10T10:00:00+00:00\t0\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\n11b5a3e531b551d1e29c9a44d91c68a1447184c7\t2026-04-09T10:00:00+00:00\t0\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\n","text-rule-trends/examples/made-up/vers/example_project-i__CONTRIBUTING.md.tsv","5952a16af749e38c05345bdbf271402ae8e29d06\t2026-04-28T10:00:00+00:00\t0\tIyBDb250cmlidXRpbmcKClRoYW5rcyBmb3IgaGVscGluZy4KCiMjIFJ1bGVzCgotIEFkZCBhIHRlc3QgZm9yIGV2ZXJ5IGJ1ZyB5b3UgZml4LgotIE5ldmVyIHNraXAgYSBmYWlsaW5nIHRlc3Qgd2l0aG91dCBhbiBpc3N1ZSBudW1iZXIuCi0gVXNlIGZvdXIgc3BhY2VzIGZvciBpbmRlbnRhdGlvbiBpbiBldmVyeSBmaWxlLgotIERvY3VtZW50IGV2ZXJ5IG5ldyBvcHRpb24gaW4gdGhlIHJlZmVyZW5jZSBwYWdlLgotIFVwZGF0ZSB0aGUgY2hhbmdlbG9nIHdoZW4geW91IGNoYW5nZSBiZWhhdmlvdXIuCi0gV3JpdGUgY29tbWl0IG1lc3NhZ2VzIGluIHRoZSBpbXBlcmF0aXZlIG1vb2QuCg==\n4bf267b59585a396085724296c2a01fdee69e740\t2026-03-28T10:00:00+00:00\t0\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\n90c908309dbe8eaccbb47799eaac9fbd725db617\t2026-02-27T10:00:00+00:00\t0\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\n","text-rule-trends/examples/made-up/vers/example_project-h__CONTRIBUTING.md.tsv","5b7750292086a3846d06efb935fbe982ca579bf0\t2026-03-17T10:00:00+00:00\t0\tIyBDb250cmlidXRpbmcKClRoYW5rcyBmb3IgaGVscGluZy4KCiMjIFJ1bGVzCgotIEFsd2F5cyBydW4gdGhlIHRlc3RzIGJlZm9yZSB5b3Ugb3BlbiBhIHB1bGwgcmVxdWVzdC4KLSBBZGQgYSB0ZXN0IGZvciBldmVyeSBidWcgeW91IGZpeC4KLSBQcmVmZXIgc21hbGwgZnVuY3Rpb25zIG92ZXIgbG9uZyBvbmVzIHdpdGggbWFueSBicmFuY2hlcy4KLSBEb2N1bWVudCBldmVyeSBuZXcgb3B0aW9uIGluIHRoZSByZWZlcmVuY2UgcGFnZS4KLSBVcGRhdGUgdGhlIGNoYW5nZWxvZyB3aGVuIHlvdSBjaGFuZ2UgYmVoYXZpb3VyLgotIE5ldmVyIHB1c2ggZGlyZWN0bHkgdG8gdGhlIG1haW4gYnJhbmNoLgotIFdyaXRlIGNvbW1pdCBtZXNzYWdlcyBpbiB0aGUgaW1wZXJhdGl2ZSBtb29kLgotIERvIG5vdCBjb21taXQgZ2VuZXJhdGVkIGZpbGVzIG9yIGxvY2FsIHNlY3JldHMuCg==\n8ef9fd76829e8bc5208f701cdc5c3b9b3c846440\t2026-02-16T10:00:00+00:00\t0\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\n0133984371a5fbaa16624e89bc7e28501a578bc0\t2026-01-15T10:00:00+00:00\t0\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\n","text-rule-trends/examples/made-up/vers/example_project-g__CONTRIBUTING.md.tsv","5b4ecc0560db45d531313f25080aab4d96b4659a\t2026-02-24T10:00:00+00:00\t0\tIyBDb250cmlidXRpbmcKClRoYW5rcyBmb3IgaGVscGluZy4KCiMjIFJ1bGVzCgotIEFsd2F5cyBydW4gdGhlIHRlc3RzIGJlZm9yZSB5b3Ugb3BlbiBhIHB1bGwgcmVxdWVzdC4KLSBBZGQgYSB0ZXN0IGZvciBldmVyeSBidWcgeW91IGZpeC4KLSBQcmVmZXIgc21hbGwgZnVuY3Rpb25zIG92ZXIgbG9uZyBvbmVzIHdpdGggbWFueSBicmFuY2hlcy4KLSBVcGRhdGUgdGhlIGNoYW5nZWxvZyB3aGVuIHlvdSBjaGFuZ2UgYmVoYXZpb3VyLgotIERvY3VtZW50IGV2ZXJ5IG5ldyBvcHRpb24gaW4gdGhlIHJlZmVyZW5jZSBwYWdlLgotIERvIG5vdCBjb21taXQgZ2VuZXJhdGVkIGZpbGVzIG9yIGxvY2FsIHNlY3JldHMuCg==\n46291100d6f39633c619de91aacb75d218c2081e\t2026-01-23T10:00:00+00:00\t0\tIyBDb250cmlidXRpbmcKClRoYW5rcyBmb3IgaGVscGluZy4KCiMjIFJ1bGVzCgotIEFsd2F5cyBydW4gdGhlIHRlc3RzIGJlZm9yZSB5b3Ugb3BlbiBhIHB1bGwgcmVxdWVzdC4KLSBBZGQgYSB0ZXN0IGZvciBldmVyeSBidWcgeW91IGZpeC4KLSBQcmVmZXIgc21hbGwgZnVuY3Rpb25zIG92ZXIgbG9uZyBvbmVzIHdpdGggbWFueSBicmFuY2hlcy4KLSBVcGRhdGUgdGhlIGNoYW5nZWxvZyB3aGVuIHlvdSBjaGFuZ2UgYmVoYXZpb3VyLgotIERvY3VtZW50IGV2ZXJ5IG5ldyBvcHRpb24gaW4gdGhlIHJlZmVyZW5jZSBwYWdlLgotIERvIG5vdCBjb21taXQgZ2VuZXJhdGVkIGZpbGVzIG9yIGxvY2FsIHNlY3JldHMuCg==\n0a3fa928fcb22f4dabeaa8a8cc187fbc9d7987a1\t2025-12-22T10:00:00+00:00\t0\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\n8b8bf17944542899744cd566c1e1f5cba6466f67\t2025-11-21T10:00:00+00:00\t0\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\n","text-rule-trends/examples/made-up/vers/example_project-f__CONTRIBUTING.md.tsv","8f1d3a63893b36de9eb0526f6c5f548c55a424a2\t2026-01-11T10:00:00+00:00\t0\tIyBDb250cmlidXRpbmcKClRoYW5rcyBmb3IgaGVscGluZy4KCiMjIFJ1bGVzCgotIE5ldmVyIHNraXAgYSBmYWlsaW5nIHRlc3Qgd2l0aG91dCBhbiBpc3N1ZSBudW1iZXIuCi0gQWRkIGEgdGVzdCBmb3IgZXZlcnkgYnVnIHlvdSBmaXguCi0gQXZvaWQgYWJicmV2aWF0aW9ucyBpbiBwdWJsaWMgbmFtZXMgYW5kIG9wdGlvbnMuCi0gUHJlZmVyIHNtYWxsIGZ1bmN0aW9ucyBvdmVyIGxvbmcgb25lcyB3aXRoIG1hbnkgYnJhbmNoZXMuCi0gVXNlIGZvdXIgc3BhY2VzIGZvciBpbmRlbnRhdGlvbiBpbiBldmVyeSBmaWxlLgotIERvY3VtZW50IGV2ZXJ5IG5ldyBvcHRpb24gaW4gdGhlIHJlZmVyZW5jZSBwYWdlLgo=\n5593a3853d8fcc7a3c33c76315f3e3c62de63121\t2025-12-10T10:00:00+00:00\t0\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\nd5cdc9c94a52142cf56a8216ec33e98bd8c5e8d3\t2025-11-09T10:00:00+00:00\t0\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\n403368e3f730ab03eca2106ced27c32143eb303a\t2025-10-08T10:00:00+00:00\t0\tIyBDb250cmlidXRpbmcKClRoYW5rcyBmb3IgaGVscGluZy4KCiMjIFJ1bGVzCgotIE5ldmVyIHNraXAgYSBmYWlsaW5nIHRlc3Qgd2l0aG91dCBhbiBpc3N1ZSBudW1iZXIuCi0gQWRkIGEgdGVzdCBmb3IgZXZlcnkgYnVnIHlvdSBmaXguCi0gQXZvaWQgYWJicmV2aWF0aW9ucyBpbiBwdWJsaWMgbmFtZXMgYW5kIG9wdGlvbnMuCi0gUHJlZmVyIHNtYWxsIGZ1bmN0aW9ucyBvdmVyIGxvbmcgb25lcyB3aXRoIG1hbnkgYnJhbmNoZXMuCi0gVXNlIGZvdXIgc3BhY2VzIGZvciBpbmRlbnRhdGlvbiBpbiBldmVyeSBmaWxlLgotIERvY3VtZW50IGV2ZXJ5IG5ldyBvcHRpb24gaW4gdGhlIHJlZmVyZW5jZSBwYWdlLgotIEtlZXAgdGhlIFJFQURNRSBzaG9ydCBhbmQgbGluayB0byB0aGUgZG9jcyBmb3IgZGV0YWlsLgotIFdyaXRlIGNvbW1pdCBtZXNzYWdlcyBpbiB0aGUgaW1wZXJhdGl2ZSBtb29kLgo=\n","text-rule-trends/examples/made-up/vers/example_project-e__CONTRIBUTING.md.tsv","f7d11281bd4e28fe7962cb841f733db19fcd597a\t2025-09-28T10:00:00+00:00\t0\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\n5310019c451e114129a9af7392223f1bd9b74587\t2025-08-28T10:00:00+00:00\t0\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\n","text-rule-trends/examples/made-up/vers/example_project-d__CONTRIBUTING.md.tsv","f7e142c7db859cea555447cb7199ea5d0041e689\t2025-08-15T10:00:00+00:00\t0\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\nccedbdd8317ab25bb7faa236617772dcd4f302d1\t2025-07-14T10:00:00+00:00\t0\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\n","text-rule-trends/examples/made-up/vers/example_project-c__CONTRIBUTING.md.tsv","713e8f7247faedc9561130012ae48fb509e237c3\t2025-06-21T10:00:00+00:00\t0\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\n1dbb6bb1c1976ae853f73c742691d3e6a842c142\t2025-05-20T10:00:00+00:00\t0\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\n","text-rule-trends/examples/made-up/vers/example_project-b__CONTRIBUTING.md.tsv","6d77416c49b8ed1e97764bfcaea794b0129ce891\t2025-07-06T10:00:00+00:00\t0\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\n0bfefbb69f5ef26f3040d9e5d20ba5d9334397f1\t2025-06-05T10:00:00+00:00\t0\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\n9f0a6b338b8aec9bb2c09003615d11a786616ac3\t2025-05-04T10:00:00+00:00\t0\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\n37f3898cbd3c639bb8b3b352004f13b5f084284d\t2025-04-03T10:00:00+00:00\t0\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\n","text-rule-trends/examples/made-up/vers/example_project-a__CONTRIBUTING.md.tsv","3545befd23ec29878d5445af160bb3ac86dbbeb3\t2025-05-13T10:00:00+00:00\t0\tIyBDb250cmlidXRpbmcKClRoYW5rcyBmb3IgaGVscGluZy4KCiMjIFJ1bGVzCgotIEFkZCBhIHRlc3QgZm9yIGV2ZXJ5IGJ1ZyB5b3UgZml4LgotIEFsd2F5cyBydW4gdGhlIHRlc3RzIGJlZm9yZSB5b3Ugb3BlbiBhIHB1bGwgcmVxdWVzdC4KLSBQcmVmZXIgc21hbGwgZnVuY3Rpb25zIG92ZXIgbG9uZyBvbmVzIHdpdGggbWFueSBicmFuY2hlcy4KLSBVc2UgZm91ciBzcGFjZXMgZm9yIGluZGVudGF0aW9uIGluIGV2ZXJ5IGZpbGUuCi0gQXZvaWQgYWJicmV2aWF0aW9ucyBpbiBwdWJsaWMgbmFtZXMgYW5kIG9wdGlvbnMuCi0gRG9jdW1lbnQgZXZlcnkgbmV3IG9wdGlvbiBpbiB0aGUgcmVmZXJlbmNlIHBhZ2UuCg==\n333172134f114e625aecb9e4b27b1c085f26025a\t2025-04-12T10:00:00+00:00\t0\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\n78b719cb6562723a9fcb45ebf648d17905dd95e9\t2025-03-11T10:00:00+00:00\t0\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\n6683d858dc3840505cec7b3005c68a0843d76e1b\t2025-02-10T10:00:00+00:00\t0\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\n","text-rule-trends/examples/made-up/themes.json","{\n \"_note\": \"Names for the 4 clusters in centroids.json beside this file (one k=4 fit of the made-up CONTRIBUTING.md rules). Valid only for labels assigned to those centroids.\",\n \"centroids_id\": \"sha256:2e25c70679e2e79d\",\n \"clusters\": {\n \"0\": \"Git: main branch, commits, secrets\",\n \"1\": \"Tests\",\n \"2\": \"Docs and changelog\",\n \"3\": \"Code style\"\n },\n \"themes\": [\n {\n \"name\": \"Code style\",\n \"clusters\": [\n 3\n ]\n },\n {\n \"name\": \"Tests\",\n \"clusters\": [\n 1\n ]\n },\n {\n \"name\": \"Docs and changelog\",\n \"clusters\": [\n 2\n ]\n },\n {\n \"name\": \"Git: main branch, commits, secrets\",\n \"clusters\": [\n 0\n ]\n }\n ]\n}\n","text-rule-trends/examples/made-up/config.json","{\n \"_note\": \"Smoke-test settings for the made-up CONTRIBUTING.md histories. Bins are half-years' worth of quarters with very few files, so min_files is 2 here; real data wants 20.\",\n \"mode\": \"cohort\",\n \"age\": 30,\n \"period\": \"quarter\",\n \"from\": \"2025Q2\",\n \"to\": \"2026Q2\",\n \"threshold\": 0.15,\n \"min_files\": 2,\n \"split\": \"2025-10-01\",\n \"kinds\": [\n \"CONTRIBUTING.md\"\n ],\n \"today\": \"2026-10-11\"\n}\n","text-rule-trends/examples/made-up/centroids.json","{\"n_init\": 6, \"seed\": 0, \"id\": \"sha256:2e25c70679e2e79d\", \"k\": 4, \"dim\": 384,\n \"centroids\": [\n [-0.00937245,-0.0029914095,-0.004157659,-0.027824366,0.014872443,-0.025626287,0.012821162,-0.052451562,0.037853543,-0.0127314925,0.017272355,0.05844389,0.05270845,-0.019352147,-0.0138978055,-0.012972865,-0.041267615,0.059541702,-0.02978619,0.010385796,0.07181034,-0.0066503654,0.068318725,0.007885078,-0.011128979,0.0059128357,-0.043687373,-0.012833889,-0.08257272,-0.1197218,-0.0139489835,-0.0405464,-0.04487587,0.0051473468,0.031518094,0.013347112,-0.0057278806,0.01643258,-0.058338355,0.017323906,0.00041567034,0.028313305,0.008899162,-0.0109813735,-0.022108087,-0.07563223,0.012374116,-0.03187331,0.08072797,0.0024326523,-0.00045504014,0.016633993,-0.037274927,0.02262415,-0.009649598,0.055235047,0.06070861,0.02946372,-0.0034311614,-0.0053353487,0.024759931,0.011435594,-0.11111287,0.06188056,0.03457985,0.024342561,0.008837083,-0.004836593,0.0043899436,0.023657126,-0.01887822,0.010325128,-0.044872403,0.10029328,0.005446872,-0.031839803,-0.015028304,0.04392657,-0.02090055,-0.009253727,-0.04492689,-0.035486653,-0.018787406,0.019771878,-0.052029394,-0.031732317,0.002163368,0.02793961,0.025046237,-0.0023373142,0.013515116,-0.09140888,0.075612605,0.018889893,-0.030308407,-0.013532882,0.0017863862,0.0029726936,-0.08758018,0.3785087,-0.043614704,0.047261395,-0.025295923,0.013928932,0.059223987,-0.008251122,0.023227096,0.001391329,-0.0259235,0.0055053583,0.014657132,-0.011934677,0.043681998,-0.042902317,0.031470943,0.03405321,0.010880333,0.029870536,-0.092372276,0.009800794,-0.03832242,0.036528587,0.036410216,0.07054993,0.03114427,-0.051381882,0.008965334,0.043624222,0.03758711,0.046242256,-0.017651476,-0.012685172,-0.006919867,0.025837138,0.008403962,-0.0021893985,-0.041066,0.015561759,0.07672973,-0.037807785,0.0010175313,0.016723618,-0.012095889,-0.019772815,-0.023602046,0.094935246,-0.021525647,0.055016536,-0.014953248,-0.021716326,-0.042458475,-0.011951329,-0.0016145135,-0.03258183,-0.021433396,-0.015483739,0.002567074,0.019992838,-0.03004945,-0.027487008,0.012248746,-0.010710435,-0.020751394,0.09265728,0.007352838,-0.03353091,-0.0076215314,-0.043834973,0.014336289,-0.01252603,-0.022650873,-0.0037427437,0.029382447,-0.022723965,0.07563404,-0.054816373,-0.010765027,-0.015717393,0.010431375,0.055186097,-0.0036385572,-0.014486508,0.022940377,-0.005758322,0.035122693,-0.017178621,-0.05951808,-0.034839775,0.0406366,0.0260605,-0.090025656,-0.00047420253,-0.026348976,-0.061839078,-0.03490972,-0.038597304,-0.00075583,0.0045453683,-0.017697204,-0.02272955,0.019441707,0.04924214,-0.03032781,0.022165004,0.023519507,0.04542617,-0.015175935,-0.04849591,0.05033668,0.0078050075,-0.072378926,-0.018664483,0.061277464,-0.04006911,-0.0644874,0.008826968,0.034279242,0.07159787,0.026182346,-0.009002443,-0.039020684,0.063111395,0.021399247,-0.3612329,-0.008085272,-0.0217074,-0.02695814,0.02199197,-0.037442252,0.014279009,-0.009837577,-0.041486315,0.023000035,0.026930124,-0.044526268,-0.024844091,0.025546052,-0.015570215,0.03606076,-0.028018478,0.048874713,-0.004073614,0.031095857,0.0036658752,-0.028477874,-0.046633977,-0.09515172,0.0007246954,0.024892855,0.16802226,-0.0025905953,0.0595403,-0.00032665386,0.061062716,0.047684953,0.024087232,-0.14755608,-0.022093046,0.03772418,-0.03883138,0.0337935,0.054922633,-0.00079847703,-0.003256574,-0.0010587673,-0.042411942,-0.104108475,0.042706076,-0.08344372,-0.05958379,0.01061481,-0.027171412,-0.010691898,0.004482833,-0.013197899,0.0666717,0.06397555,0.032217808,-0.056039896,-0.013437512,0.00992869,0.021784522,0.0063591697,-0.014027406,-0.023403194,0.035053518,-0.03292476,0.002936119,-0.0034598468,-0.012779367,0.027969113,0.058428857,0.016792173,-0.0076702368,0.1354448,0.008010893,-0.088306814,0.0028661923,0.033256,-0.0405116,-0.008934119,-0.083088376,-0.03621356,-0.005376729,0.021023428,0.06361798,0.00078189996,-0.02458382,0.06334421,0.04485816,-0.0020086856,0.019811403,-0.0053502265,-0.017100884,0.000044247085,-0.0260583,0.033317108,0.104152754,-0.016740447,-0.2842223,-0.030426053,-0.010690118,-0.008194267,0.016415698,0.038626213,0.013083365,-0.06557467,-0.092236675,0.06945641,-0.008518564,0.04512242,-0.032380063,-0.043582425,0.04062892,0.016126238,0.08234648,0.025265582,0.033600423,-0.04504981,0.013522106,-0.012505811,0.20510022,-0.0124264,-0.025498206,0.01627119,-0.022336928,0.06413923,0.023656854,0.025231212,-0.006674338,-0.017896697,0.015798246,-0.0026350375,0.0119498335,-0.013630904,0.01644214,-0.017331101,0.005796313,-0.026082018,-0.060903177,0.02557889,0.032121524,-0.013771643,0.080072045,-0.0333732,-0.0129347425,0.021340087,0.016973339,-0.009478991,-0.017796645,0.0047385087,-0.023091896,0.037897326,0.040751148,0.04076797,-0.0047650114,-0.032078896,-0.052049886,0.015684394,0.0278583,-0.015881285,0.018925352,0.049613312,-0.0144294845],\n [-0.089495204,0.04951908,0.07425585,-0.011208565,0.0034947663,0.038553435,0.045140542,0.008466216,-0.020199798,-0.03373217,-0.0025716226,0.016999457,0.021239486,0.059122335,-0.02643007,-0.0039811134,-0.027743023,0.02793222,-0.024440642,-0.019287009,0.014184449,0.0042309854,0.040751547,0.020869186,-0.03332161,0.016329981,0.008763494,-0.052281894,-0.05648059,-0.14581949,0.0138732325,-0.060156226,-0.020641364,-0.03298043,-0.0043115355,-0.018463781,-0.036243122,0.034145944,-0.00680261,-0.008970902,0.051726058,0.057884544,-0.03797427,-0.04710039,-0.04464734,-0.043087948,0.05587921,-0.00896559,0.06921598,0.0037846304,-0.020791626,-0.05439293,0.026570458,-0.012741258,0.0015014072,0.0021901473,0.047778763,0.05615166,0.005057177,0.028401943,0.018673766,-0.0036325247,-0.09153469,0.07194571,0.04643396,0.00404615,-0.01580694,-0.012754649,0.029623056,0.0827415,-0.051129196,0.010095198,0.007502952,0.12481295,0.024586719,0.020607222,-0.029427519,0.000020752543,0.005731401,-0.011759753,-0.032908455,-0.03857943,0.010583052,-0.0021838965,0.015109352,-0.033751976,0.03805077,0.042337872,0.07120986,0.024654377,0.03426433,-0.0041860538,0.03697194,0.029891947,-0.010496831,0.0026448467,-0.004762455,-0.053761948,-0.02623015,0.41833827,-0.0146019235,0.01916538,-0.038041994,0.03707334,-0.005616179,-0.012075243,-0.015331708,0.011200432,-0.058076993,-0.0035032842,0.05205209,0.0005734892,0.09919907,-0.04872239,-0.018810263,0.055356793,0.009729047,-0.01573128,-0.03028236,-0.018439023,0.010579133,0.04818943,0.05070029,-0.03390762,0.035526883,-0.024629498,0.042352322,0.08239029,-0.040054478,0.0033068943,0.030526713,-0.049949367,-0.03807742,-0.00048054548,-0.007111432,0.016699474,-0.012429638,0.029337628,0.030104794,0.017563632,-0.025521487,-0.011526223,0.008470894,-0.046664923,-0.043053783,0.052104395,-0.0069011245,0.057283547,0.0018083291,-0.12331424,0.029935963,0.0064835837,-0.0134406695,-0.052045338,0.013850433,0.008110819,0.05322124,0.039829932,-0.0012629305,-0.00788727,-0.03924201,0.02717886,-0.053714506,0.024800245,-0.016731003,-0.061436214,-0.012017112,-0.022841629,-0.0064678937,-0.014485089,-0.009248493,0.034502324,-0.03129,-0.081952676,0.02668194,-0.03557625,0.007407181,0.032162644,0.014334858,0.0076508475,0.034903623,0.006391783,-0.018761557,-0.007000027,-0.0000884816,-0.041428037,-0.021331456,-0.009003584,0.019837359,-0.015723709,-0.021863958,-0.020088213,-0.0067986255,-0.0044078073,0.0038651593,-0.053588405,-0.04200801,-0.014680361,-0.022805741,-0.055535726,0.026091479,0.011984245,0.031657893,0.0022696892,0.023587454,0.07453759,0.0073553757,-0.063671626,0.08749354,0.024220815,-0.07145736,-0.029192507,0.11064528,0.0086902045,-0.034106854,0.005597543,-0.0071928785,0.029055234,0.0069394424,0.018180445,0.030818716,0.08091793,0.061646596,-0.34437498,-0.030435331,-0.014971005,0.032188077,0.007367687,-0.015100414,0.006860625,0.01038642,-0.007393869,0.01987282,0.078569,0.035064314,0.003542388,-0.059896514,0.007131672,-0.0032716352,-0.008641555,0.0057681478,-0.05881163,0.045454215,-0.032458976,-0.03906742,0.0062166573,-0.10703419,0.036208436,-0.020450393,0.16024062,-0.006728231,0.009132719,-0.030865232,-0.0050642635,0.00816569,0.021808686,-0.12920673,0.051784463,0.043808095,-0.071410626,0.065622956,0.01083293,-0.05884633,0.025733901,0.026852436,-0.025614023,-0.13224016,-0.012199966,-0.027523037,-0.020675471,-0.010853031,-0.058956448,-0.057498615,-0.02906537,-0.016964175,0.05389172,0.03021506,0.016635181,-0.069220744,-0.02874754,-0.024591148,0.014199083,-0.048656598,-0.004664231,0.04070601,-0.01679785,-0.011166486,0.054048046,-0.0040319446,-0.00955942,-0.038482223,0.06585841,0.008914155,-0.020598331,0.03555914,0.0017249582,-0.03803948,-0.03174666,-0.020482546,0.013253714,-0.040043686,-0.020404657,-0.006070975,0.0008040956,-0.01699807,0.027126223,0.030193308,0.009584524,0.050498385,-0.004307506,-0.0017854206,0.06707375,0.014338084,-0.050182056,0.022868576,-0.0029511421,-0.016412944,0.041832477,0.02739911,-0.23743118,-0.011567712,-0.0363925,-0.007876031,-0.020317638,0.041477602,-0.01336374,0.0008013951,-0.02394395,0.014161633,-0.007736382,0.06023107,-0.007461616,-0.03803263,0.0039569535,0.060906302,0.098080695,-0.031179769,0.040308096,-0.09798698,0.029465549,0.011098835,0.20892745,0.019666962,0.009739388,0.033457313,0.05459977,0.017645393,0.02181859,0.0015547898,0.027477881,-0.050683755,0.019567737,0.01382606,0.015319701,0.024312789,-0.04158637,0.008482448,-0.032792863,-0.02541984,-0.024708407,0.003202039,-0.00085070357,0.021722017,0.08966548,-0.055033274,-0.043000724,-0.025363289,-0.06434536,0.01619669,-0.011857167,-0.08812704,0.0016044853,-0.01031505,0.008456532,0.04289471,0.007978695,0.025949592,-0.02465071,-0.03915572,0.037589222,-0.012414894,0.08758092,0.038091063,-0.011827875],\n [-0.070929796,-0.0017145738,0.03515555,-0.013137637,-0.014899909,0.023396859,-0.032182526,0.037581727,0.033477098,0.02441046,0.011226018,0.07269005,-0.016929135,0.04366064,-0.023772432,-0.006082228,-0.015593478,-0.02205022,-0.020001039,0.018271457,0.055422883,0.01731845,-0.014283415,0.015331771,-0.024441803,0.025443308,-0.032022536,-0.04092694,-0.051923674,-0.12206901,-0.029798687,-0.008315964,-0.012658649,0.033146344,-0.009072528,-0.007121336,-0.026734428,0.0795499,0.03809517,0.023222424,0.026375823,0.041927263,-0.016173707,0.012433173,0.005122552,-0.031202361,0.04931518,-0.0217689,0.0704498,-0.0046823374,-0.05792852,-0.03181267,-0.029825764,-0.039276738,-0.01940594,0.02237075,0.0749016,0.007290965,-0.01750191,0.016850727,0.034413945,0.08852194,-0.17149654,0.09962453,-0.024027953,0.026896602,-0.0740235,0.016027378,0.008988,-0.010190572,-0.040989537,0.0007024038,0.0008382741,0.08470571,0.038209293,-0.056635816,-0.0026512374,0.008211346,-0.050766844,0.0019181997,-0.026118847,-0.016486581,-0.0053511714,-0.008792285,-0.043161966,-0.0068166647,0.047179352,0.020685762,0.060847435,-0.0066612167,0.044158045,-0.13636018,0.013821536,0.03303709,-0.051637214,-0.015287858,0.012520341,-0.02061694,0.014580437,0.39115188,-0.06572707,-0.01782877,0.033507418,-0.016779387,-0.0025873645,-0.013750551,0.015339644,-0.024590509,-0.03233716,0.021099485,-0.013887513,-0.025416749,0.0054932986,-0.06460133,0.026866963,0.019787006,0.007229501,-0.009708766,0.00031270046,-0.046896018,0.017822973,0.006788839,0.015165831,0.06221314,0.017377285,-0.0029068764,0.08559589,0.032111507,-0.0049725305,0.052053694,0.054985512,0.035950117,-0.057794414,0.009659274,0.04387619,-0.033465162,0.007020401,-0.041793585,0.026706481,0.0016728174,0.019040534,-0.0060453643,-0.049099468,-0.07654208,-0.074386746,0.1771004,-0.014804926,0.0021574346,-0.0141672,-0.07278306,0.038458083,0.03899018,0.063383095,-0.017704306,0.0033694815,-0.0042911503,0.061755598,-0.004805996,-0.006989453,-0.05302719,-0.02281322,0.010759785,-0.0065381867,0.083079405,-0.020972367,-0.09764888,-0.023397911,-0.014785888,-0.0030529038,-0.027061526,-0.020565223,0.040308267,-0.002133118,0.045226064,0.110312596,0.049468916,-0.013584222,0.020228839,0.06361909,-0.030576745,0.009484674,-0.07290955,-0.0719834,0.02368261,-0.0134981675,0.037689023,-0.031915504,-0.061839186,0.017970549,0.06292949,-0.008181893,0.024859114,0.047185995,-0.027671346,-0.027706541,-0.050379533,-0.07934387,-0.046047583,0.003606221,-0.032190233,-0.0018139293,0.0049047,-0.04937974,0.047033425,0.05219104,0.0011469036,-0.020430531,-0.019976385,0.1051389,-0.031219432,-0.12805958,0.01317128,0.09136219,-0.007503066,-0.0048330207,-0.018772017,-0.017458413,0.037785877,-0.010492869,0.0068947407,0.053280544,0.025995895,0.034389902,-0.2871726,0.017598037,-0.02332459,-0.0127150435,-0.016253032,-0.0150131,0.057343315,-0.016065624,-0.029943366,0.048375756,0.073634714,-0.018118188,0.007600334,-0.032938056,-0.038821477,-0.042042356,0.025845058,-0.008030089,-0.03129157,0.04004038,0.038106717,-0.005924104,-0.023988774,-0.095210224,-0.0019268397,-0.0155748455,0.1349361,-0.016171703,-0.01065396,-0.022170912,0.047353942,0.04699718,-0.022607164,-0.118830666,-0.019590832,0.09237171,-0.053574294,0.061637565,0.006556952,-0.053993825,0.059441805,0.028877668,0.025453499,-0.06090422,0.025934722,-0.0274799,-0.04328977,-0.03435058,-0.0055816025,0.004979116,-0.016310843,-0.005277714,0.010424407,0.0092189135,-0.0060280934,-0.012290152,0.000006135228,-0.007619952,-0.0029680384,-0.0442781,-0.017924657,0.012729068,0.0022145621,-0.028901396,0.035019007,-0.004234324,0.007061109,-0.02672349,0.0070977765,-0.02140246,-0.02380631,-0.020748042,0.021325158,0.01979602,-0.00918371,0.019218897,-0.0068379925,-0.0072968835,-0.051697586,-0.022158138,0.0061737257,-0.0018420127,0.035991196,0.08146247,-0.02775454,0.04928743,-0.006918696,-0.039856676,0.033327516,-0.006784337,-0.042580776,-0.02675392,-0.02890656,0.014902444,0.03581292,-0.010457975,-0.25803363,0.02816355,0.01163043,0.044037033,0.00077176717,0.076403335,0.031988397,-0.023015935,0.026368042,0.041910462,-0.023735933,0.055844482,-0.07234806,-0.056496933,-0.009765955,0.060958445,0.10544508,-0.012315726,0.0055999747,-0.02562619,0.018933823,-0.0016092842,0.19532989,0.03265119,-0.0028926458,0.022424648,0.00341616,0.021075448,0.0022406501,0.02335164,0.029012725,-0.022338962,0.04525491,-0.008863917,-0.03687438,0.018566404,-0.026857173,-0.023305427,0.013941531,-0.04021398,0.015011151,-0.014590671,-0.029415505,-0.020735968,0.065777026,-0.05422613,-0.027941167,-0.022244817,-0.09301792,0.04640722,-0.067279555,-0.02193891,0.00094477925,0.021406429,0.075722985,0.033156697,0.026962081,-0.020461457,-0.00825253,-0.018529302,-0.0045998893,-0.0060046394,0.02146897,0.04094022,0.0339269],\n [-0.03220128,-0.029432574,0.027144033,-0.04947651,0.0058935527,-0.0093420865,-0.034097288,-0.0062302724,0.047318246,-0.007898661,-0.018031016,0.0140767265,0.023108415,-0.005238245,0.028975643,-0.008981572,0.0023778987,0.04246414,-0.0745192,0.0074108783,0.07753,-0.009939742,0.036037203,-0.03402365,-0.029197218,0.03665176,-0.006150151,-0.028517116,-0.031260584,-0.17715646,0.033975184,0.035510637,0.04524877,0.027246704,-0.046113513,0.014880181,-0.01584114,0.023991618,-0.071058564,0.010618262,0.062942274,0.03063819,0.0045494204,0.028018543,-0.060983527,-0.0940654,-0.061926853,-0.020032302,0.022364143,0.030307122,-0.039436303,-0.03566812,0.00901424,0.034379326,-0.022587271,0.010370586,0.0150409555,0.035587717,-0.012444207,-0.03908138,0.06540482,0.030013239,-0.10014952,0.08549877,0.004419259,-0.026273929,-0.006943937,-0.017677175,0.006751784,0.07550037,-0.050287243,0.017313061,-0.038295165,0.10359182,-0.007918741,-0.05133158,-0.0046835127,0.021140594,-0.023312246,0.013121117,-0.01579663,-0.05197574,-0.057355363,0.003125099,-0.06393928,-0.020130062,-0.020714749,-0.02199019,0.03528895,0.03375951,-0.009777912,-0.028008826,0.04648431,0.030295609,-0.047903392,-0.06590424,0.013749785,0.02287219,-0.02024242,0.38747376,-0.035570275,0.024995284,0.031070635,-0.028304933,-0.01055569,0.04742541,-0.0105339335,0.0034680255,-0.005476796,-0.022223325,-0.014559592,-0.06592574,0.020260895,0.0011014071,-0.031715967,-0.021652937,-0.01096091,-0.025009649,-0.0081950575,0.039768018,-0.045273136,-0.004183331,-0.043067757,0.025568387,0.019124296,-0.0082467925,0.00074574916,0.06522813,0.027805842,0.0530028,0.032096304,0.018631168,0.05243715,-0.0021898455,0.0069323843,0.010570708,0.015906354,-0.0013107725,0.049414046,-0.03787631,-0.01691669,-0.022181654,0.0017500933,-0.031997226,0.0033198474,0.10654066,-0.06956426,-0.0018172035,0.02366725,-0.045393214,0.011099229,0.071073756,0.012808508,-0.051945556,0.019830482,0.031143265,0.008925068,-0.019768277,-0.03204752,0.00048669273,-0.03857761,0.057908952,0.0019075223,0.05123157,-0.021168409,-0.035427604,-0.033651482,0.021712434,-0.03243607,-0.019181546,0.008191124,-0.021599503,0.021189416,0.02665219,0.093797155,-0.0076598423,-0.071688294,-0.008898139,0.016368063,0.07032759,0.05347055,0.005132844,-0.000046288133,0.023062522,0.026221216,0.020715954,-0.043579962,0.0043859244,0.03638438,0.026432313,-0.04051008,0.060951862,-0.049895473,-0.012465259,-0.061976545,-0.060080186,0.010163489,-0.000814036,-0.04592147,0.02050062,0.029821139,0.019964108,-0.050962035,0.0072903363,0.036341768,-0.03608476,0.0349778,-0.04573764,0.034424204,0.0008420406,-0.07393494,-0.026229715,0.09699801,-0.04354563,-0.028568871,-0.055714704,0.018680403,0.024959153,0.05056724,0.0020132733,0.010262489,-0.021339694,-0.010922527,-0.3556777,0.024013663,0.0015654962,-0.038558498,0.052992493,-0.06677115,-0.0060268166,0.061937515,-0.05099366,0.014896889,0.006388496,-0.060752116,-0.008206272,0.005719976,-0.04468499,0.052446205,-0.00060254725,-0.004930764,-0.06823171,0.0064955605,0.05825896,0.037178498,0.0013243239,-0.042423114,0.0109422505,0.018557524,0.1804236,-0.032737423,0.037091494,0.026630666,0.09826146,0.01435546,0.0039822333,-0.05601835,0.013567514,0.06402381,-0.0071678944,0.058664225,0.018736491,-0.0063439086,0.033068676,0.024507621,-0.009519264,-0.041931853,-0.009755541,-0.039447054,-0.031358086,-0.0613942,-0.009388329,-0.014017754,-0.03810867,-0.05168645,0.03845141,0.016995164,0.017969476,-0.027628195,-0.024485908,-0.010138164,-0.0026830111,0.013356937,0.0033422709,-0.06175607,-0.008085909,-0.011008183,0.0109200245,0.013959343,-0.0070002945,0.012515072,-0.006585949,-0.029263379,-0.036916316,0.03457865,-0.034659896,-0.04424275,0.017358543,0.011917324,-0.009444115,0.108842686,-0.02095262,-0.05994754,0.033350464,0.015782375,-0.0023176728,-0.020009192,0.017567586,0.004517228,-0.0056561185,0.033515684,0.057687428,-0.026646044,0.032062255,0.043210875,-0.008567531,0.07501279,0.052940335,-0.0022104376,-0.2899056,0.009893856,-0.0041691572,0.00810704,-0.027037542,0.027053492,-0.04580092,-0.0797439,0.021736873,-0.029319508,-0.012398258,0.049616687,-0.014459076,-0.06164577,0.039153177,0.007168662,0.06831914,-0.013162054,0.04803639,-0.068883665,0.07413286,0.009373121,0.2189224,-0.032952417,0.036878277,0.044615146,-0.0024079806,-0.01445946,-0.021129264,0.03180072,0.047423992,-0.017809328,0.06059423,-0.03076188,0.015767578,0.07081767,-0.026143538,-0.016350193,0.000083016144,-0.0014837342,-0.037601378,0.006876858,-0.0736585,-0.025016032,0.07223769,-0.017569395,0.0149451895,-0.0034171087,0.004955423,0.007764619,-0.013954625,0.027629558,-0.03461301,0.007885398,0.043427743,0.029928429,-0.024712197,0.01510591,-0.0069634668,0.018343521,-0.03504719,-0.07286319,0.022522489,0.05335717,0.0388483]\n ]\n}\n","text-rule-trends/examples/made-up/NOTICE","These histories are made up for this skill's smoke test: twelve imaginary projects, each\nstarting a CONTRIBUTING.md and editing it a few times. Later projects write fewer style rules\nand more git rules, so the trend has a known shape. Same MIT licence as the skill.\nFormat: one row per version, newest first: sha author date 0 base64 body.\n","text-rule-trends/SKILL.md","# Text rule trends\n\nTurns per-file git histories into numbers about what those files say, and how that changed.\n\nIt extracts line-level items (by default \"rules\": imperative or normative lines) from every version\nof each file and follows each item's life. It reads each file at an age you choose (or at the end of\nevery period), embeds the items with a sentence model (bge-small by default), clusters them with\nyour own k, maps clusters to themes from your JSON, bins by month, quarter or year, labels each\ntheme rise, fall or steady with your threshold, bootstraps a before/after comparison around any\ndate, and counts one file kind's share of new files against another. The extractor can be swapped\nfor a regex or your own plugin function. It reads histories from `github-file-history` and writes\nthe JSON that `tufte-charts` and `watercolor-data-art` draw.\n\nExample output (`analyze.py` on the made-up smoke test below):\n\n```\n== themes, share of a file's items 2025Q2 -> 2026Q2\n rise Git: main branch, commits, secrets 25.0% -> 37.5%\n rise Docs and changelog 12.5% -> 20.8%\n fall Tests 37.5% -> 29.2%\n fall Code style 25.0% -> 12.5%\n```\n\nUse it when:\n- you collected many versions of one kind of file (with `github-file-history`) and want to know\n which kinds of lines appear, grow or disappear: rules in CLAUDE.md or `.cursor/rules`,\n sections in READMEs, dependencies in package.json, steps in CI files\n- you want input for `tufte-charts` or `watercolor-data-art`\n\n## Requirements\n\nPython 3.8+ and `pip install -r requirements.txt` (numpy, scikit-learn, fastembed). fastembed\ndownloads the embedding model the first time (bge-small is about 70 MB). Use `--cache-dir` to\nchoose where. `sentence-transformers` is an optional second backend.\n\n## Inputs\n\n- One folder per sample, holding `vers/` as written by `github-file-history`: one `.tsv.gz`\n or `.tsv` per file, one row per commit (sha, author date, Claude co-author 0/1, base64\n body). A `manifest.tsv` beside it, if present, maps keys to paths.\n- A themes file (your JSON) that names clusters and groups them into themes, bottom to top:\n `{\"centroids_id\": \"sha256:...\", \"clusters\": {\"0\": \"name\", ...}, \"themes\": [{\"name\": \"...\",\n \"clusters\": [0, 3]}, ...]}`. Without one, every cluster is its own theme.\n `centroids_id` is the `\"id\"` printed in the `centroids.json` the names were written for. Cluster\n numbers mean nothing outside that one set of centroids, so `analyze.py` refuses a themes file\n whose `centroids_id` differs from the labels' (`--force-themes` overrides), and refuses samples\n labelled against different centroids.\n- Optionally a config file with the analysis settings (see `examples/*/config.json`).\n\n## Steps\n\nRun the scripts by their path, and write outputs to a folder of your own, never inside the skill\nfolder. Below, `SKILL` is this skill's folder and `P`, `R` are your sample folders.\n\n```bash\n# 1. items and their lives, per sample -> P/rules.json\npython3 $SKILL/scripts/extract.py $P # the default \"rules\" extractor\npython3 $SKILL/scripts/extract.py $P --extractor regex --regex '^#{1,3}\\s+(.+)' # Markdown headings\npython3 $SKILL/scripts/extract.py $P --extractor plugin --plugin my_items.py:items\n\n# 2. embeddings, per sample -> P/emb.npy (a local working file)\npython3 $SKILL/scripts/embed.py $P [--model BAAI/bge-small-en-v1.5] [--cache-dir DIR]\n\n# 3. clusters: fit on one sample, assign the others to its centroids\npython3 $SKILL/scripts/cluster.py fit $P --k 40 > clusters.txt # -> P/labels.json, P/centroids.json\npython3 $SKILL/scripts/cluster.py assign $R --centroids $P/centroids.json\n\n# 4. read clusters.txt, name the clusters and group them into themes.json,\n# with \"centroids_id\" set to the \"id\" in P/centroids.json\n\n# 3b. or reuse someone's named clusters: assign EVERY sample to their centroids.json\npython3 $SKILL/scripts/cluster.py assign $P --centroids their/centroids.json\n\n# 5. trends -> OUT/*.json\npython3 $SKILL/scripts/analyze.py --sample a=$P --sample b=$R --themes themes.json \\\n --config config.json --out OUT\n```\n\n### Extractors (`extract.py`)\n\n| `--extractor` | what is an item | default words |\n|---|---|---|\n| `rules` | a list item, or a line starting with a capital letter, whose text (after an optional bold label) starts with an imperative or normative word: always, never, use, run, avoid, must, should, ... Fenced code, headings and table rows are skipped | 4 to 45 |\n| `lines` | every non-blank line outside fenced code, list markers removed | 1 to 45 |\n| `regex` | lines matching `--regex`; the item is group 1 if there is one, else the whole match | 1 to 45 |\n| `plugin` | whatever your function returns: `--plugin FILE.py:FUNC`, where `FUNC(body)` returns a list of strings | 1 to 45 |\n\nOther settings: `--min-words`, `--max-words`, `--match` (default 0.6: an item keeps its identity\nthrough a rewording when its words overlap that much, by token Jaccard, with an unmatched item\nof the previous version), `--kind-map` (a JSON `{\"regex\": \"label\"}` matched against each file's\npath, to group file names into kinds; the default kind is the file name), and `--order`.\n\n**Version order.** Versions are replayed oldest first. `--order datetime` (default) sorts by the\nparsed author date, so mixed UTC offsets come out in the right order. `--order string` sorts by\nthe date text, which some older runs used; pass it to reproduce one of those.\n\n### Analysis settings (`analyze.py`, or keys in `--config`)\n\n| setting | default | meaning |\n|---|---|---|\n| `mode` | `cohort` | `cohort`: bin files by creation period, read each at creation + `age`. `snapshot`: in every period, read every live file as it stood on the period's last day |\n| `age` | 30 | days after creation |\n| `created` | `first-item` | a file's creation date: its earliest item, or `first-version` (its first commit) |\n| `period` | `quarter` | `quarter`, `month` or `year` |\n| `periods` | all with data | the bins to report |\n| `from`, `to` | first, last bin | the two bins compared for rise and fall |\n| `threshold` | 0.15 | relative change in a theme's share that counts as rise or fall. A theme that starts at 0 and grows is `rise`. If `from` or `to` is an empty bin, its shares are `null` and every theme is `too few` |\n| `min_files` | 20 | if a compared bin has fewer files, every theme is `too few`; a window side with fewer is flagged too few |\n| `before`, `after` | none | windows `A:B` and `C:D` by creation date, `[A, B)` and `[C, D)` |\n| `split` | none | one date instead: before = every file created earlier, after = created on it or later |\n| `kinds` | all, most files first | kinds reported separately, in this order (the bootstrap draws follow it) |\n| `share` | none | `A:B`: kind A's share of new files among A and B, per period |\n| `boot`, `seed` | 3000, 3 | bootstrap resamples and numpy `default_rng` seed |\n| `today` | **the current date** | files younger than `age` on this day are left out, and share.json leaves out files first committed after it. Set it to make a run repeatable |\n\nWindows compare files by when they were created, each read at creation + `age`, in both modes. In\nsnapshot mode the bins are calendar periods, but the windows still work this way.\n\n## Outputs\n\n| file | schema |\n|---|---|\n| `/rules.json` | `{nfiles, nhistory, life: [{kind, born, died, text, norm, repo}], files: {key: {kind, first, deleted}}, co_month, settings}`. `nhistory` counts history files found, `nfiles` those with a version. `repo` is the history file's name (one file in one repo), kept under that name for compatibility |\n| `/emb.npy` | float32 `[items, dim]`, same order as `life`. A local working file |\n| `/labels.json` | `{k, labels, centroids_id}` |\n| `/centroids.json` | `{k, dim, centroids: [[...]]}`, unit length, plain text. Values are the shortest decimals that read back as the same float32, so assigning with the JSON gives the same labels as a `.npy`. `cluster.py convert` turns one into the other |\n| `OUT/pooled.json` | `{Q, nfiles, tot, med, pf, files_with_rules, overlap}`: per period, files, mean and median items per file, mean items per file in each cluster. An empty bin is `null`, never 0 |\n| `OUT/themes.json` | `{q, nfiles, tot, med, from_q, to_q, threshold, groups: [{name, kind, v, share_from, share_to}], rest, files_with_items, items, samples: {name: {history_files, files, files_with_items, items}}}`. `kind` is `rise`, `fall`, `steady` or `too few`. `share_from` and `share_to` are **fractions** (0.181 = 18.1%). Empty bins are `null`. The input to both chart skills |\n| `OUT/compare.json` | per sample and kind: `periods` (n, mean, median, pf per period, whatever the unit) and `window` (means, medians, difference, bootstrap 95% CI, and `too_few`) |\n| `OUT/share.json` | per sample, with `share`: `{Q, files, history_files, : [new files per period], share_of, share_vs, share}`. `share` is in **percent** (87.2 = 87.2%). Counts every file by its first commit, with or without items, over `periods` if set (else every period), leaving out files first committed after `today` |\n\n## Examples\n\n### Smoke test: made-up CONTRIBUTING.md rules (`examples/made-up/`)\n\n`vers/` holds twelve made-up projects, each starting a CONTRIBUTING.md and editing it a few times.\nLater projects write fewer style rules and more git rules, so the trend has a known shape.\n`centroids.json`, `themes.json` and `config.json` beside it name one k=4 fit. Under a minute after\nthe model download:\n\n```bash\nmkdir -p /tmp/trt && cp -r $SKILL/examples/made-up/vers /tmp/trt/\npython3 $SKILL/scripts/extract.py /tmp/trt # files 12 with items 12 items 96\npython3 $SKILL/scripts/embed.py /tmp/trt # embedded 96 items\npython3 $SKILL/scripts/cluster.py assign /tmp/trt --centroids $SKILL/examples/made-up/centroids.json\npython3 $SKILL/scripts/analyze.py --config $SKILL/examples/made-up/config.json \\\n --sample made-up=/tmp/trt --themes $SKILL/examples/made-up/themes.json --out /tmp/trt/out\n```\n\nExpected: \"Git: main branch, commits, secrets\" rises (25.0% → 37.5% of a file's rules, 2025Q2 →\n2026Q2), \"Docs and changelog\" rises (12.5% → 20.8%), \"Tests\" falls (37.5% → 29.2%) and \"Code\nstyle\" falls (25.0% → 12.5%). The split at\n2025-10-01 compares 5 files with 7. The config uses `min_files: 2` because the bins are tiny; real\ndata wants the default 20.\n\n### README section headings (`examples/readme-sections/`)\n\nThe README histories of the five repos in `github-file-history/examples/readme-package`, read as\nMarkdown headings, in snapshot mode by year:\n\n```bash\npython3 $SKILL/scripts/extract.py $D --extractor regex --regex '^#{1,3}\\s+(.+)' \\\n --kind-map '{\"(?i)^readme\": \"README\"}' # files 12 with items 7 items 151\npython3 $SKILL/scripts/embed.py $D\npython3 $SKILL/scripts/cluster.py fit $D --k 6 # then name the clusters: themes.json\npython3 $SKILL/scripts/analyze.py --config $SKILL/examples/readme-sections/config.json \\\n --sample five=$D --themes $SKILL/examples/readme-sections/themes.json --out OUT\n```\n\nOn the data collected 2026-10-11, headings per README went from 13.0 (2016) to 15.8 (2025), and\n\"API reference entries\" rose from 35.9% to 45.6% of them. `themes.json` names that one k=6 fit;\na new collection or fit needs new names. `package_deps.py` is a plugin example: each dependency\nin a package.json becomes an item (93 items from the four package.json histories).\n\n**Worked example.** These skills were first written for the agentleFS article \"CLAUDE.md best\npractices\" (https://agentlefs.com/blog/claude-md-best-practices), which studied 5,786 CLAUDE.md and\nAGENTS.md files. Its data, settings and step-by-step recipe live outside the skills, in the\n`article-claude-md-best-practices` folder (`RECIPE.md` and `data/`).\n\n## Limits and caveats\n\n- **The extractors are heuristics.** The rules extractor counts imperative lines in English and\n misses rules written as prose paragraphs or tables. Check a sample of `rules.json` by eye.\n- **Clusters are a model's grouping; themes are a person's.** Cluster numbers depend on the\n embeddings, the sklearn version and the data. A fit on new data gives new clusters, even with\n the same k and seed: in one case, a fit on a fresh collection of the same kind of files matched\n the older centroids at 0 of 40 positions. So to reuse named themes on new data, assign every\n sample to the centroids they were named for. `cluster.py fit --match-to OLD.json` renumbers a new\n fit to line up with old centroids (Hungarian matching on cosine) and reports how close they are,\n but its clusters are still new ones, and analyze.py will refuse the old themes for them. Labels\n from the fit itself and labels assigned to its centroids differ for a few items (about 1%).\n- **Small numbers.** A bin or window side with fewer than about 20 files is noise. `analyze.py`\n marks those `too few`. Look at `nfiles` before you read a bin.\n- **Dates show timing, not cause.** A before/after split on a date cannot separate that date from\n anything else that happened then.\n- **Sample bias** carries over from collection. Report samples separately before you pool them.\n- **Licences.** `rules.json` quotes the files. Only publish it, or quotes from it, for repos whose\n licence allows that.\n\n## Credits\n\n- **bge-small-en-v1.5** by BAAI (Beijing Academy of Artificial Intelligence), the default\n embedding model, https://huggingface.co/BAAI/bge-small-en-v1.5 . MIT licence per its model card\n (checked 2026-10-11). fastembed downloads the quantized ONNX port\n `Qdrant/bge-small-en-v1.5-onnx-Q`, also MIT per its model card. The model is downloaded, not\n bundled; `centroids.json` files hold k-means centres computed from its vectors.\n- **fastembed** by Qdrant (Apache-2.0, per the package metadata of 0.9.0): the embedding runtime.\n- **scikit-learn** (BSD-3-Clause): `KMeans` for `cluster.py fit`. **numpy** (BSD-3-Clause) for\n the vectors and the bootstrap; **SciPy** (BSD-3-Clause) for the optional Hungarian matching in\n `cluster.py fit --match-to`. All are installed by you, not bundled.\n\n","text-rule-trends/LICENSE","github-file-history/scripts/make_list.py","#!/usr/bin/env python3\n\"\"\"Build the list collect.sh reads: one owner/repo/PATTERN per line.\n\nPick one source of repos, then say which paths you want with --path (repeatable; a literal\npath or a glob such as '.cursor/rules/*.mdc'). Every repo is crossed with every --path.\n\nSources:\n file FILE one owner/repo per line (a GitHub URL or an owner/repo/path line is\n cut down to owner/repo). Lines starting with # are skipped.\n search FILE a GitHub search export: the JSON of `gh search repos ... --json fullName`,\n a saved REST response from /search/repositories ({\"items\": [...]}), a\n JSON list of names, or a CSV with a full_name, fullName, nameWithOwner,\n repo_name or repo column.\n census FILE --n N a random draw of N repos from any census list (CSV with a repo_name\n column, gzipped or not, or one owner/repo per line), with\n random.Random(--seed).sample. --sampled writes the drawn repos, so the\n denominator is on record.\n catalog the agentleFS public catalog sitemap (https://agentlefs.com/context/sitemap.xml).\n It lists agent instruction files (AGENTS.md, CLAUDE.md, SKILL.md, llms.txt)\n in public repos with 1,000+ stars whose licence allows redistribution.\n Without --path it lists the catalog's own files (root files only, unless\n --any-depth; --names filters by file name). With --path it is just a\n source of repos.\n\nOptions:\n --path PATTERN a path or glob to collect in every repo (repeatable)\n --probe keep only (repo, path) pairs whose file exists today on the default\n branch, by a HEAD request to raw.githubusercontent.com (no token, no API\n rate limit). Literal paths only; globs are kept as they are.\n --jobs 16 probes at once. Non-404 errors are retried 4 times, then reported on\n stderr as \"ERR

  \".\n  --limit N              stop after N output lines\n\nExamples:\n  python3 make_list.py file repos.txt --path README.md --path package.json > list.txt\n  python3 make_list.py census census.csv.gz --n 30000 --seed 1 --path CLAUDE.md --path AGENTS.md \\\\\n      --probe --sampled drawn.txt > random.txt\n  python3 make_list.py catalog --names CLAUDE.md AGENTS.md > popular.txt\n  python3 make_list.py search repos.json --path '.cursor/rules/*.mdc' > list.txt\n\"\"\"\nimport argparse\nimport concurrent.futures as cf\nimport csv\nimport gzip\nimport io\nimport json\nimport random\nimport re\nimport sys\nimport time\nimport urllib.error\nimport urllib.request\n\nSITEMAP = \"https://agentlefs.com/context/sitemap.xml\"\nUA = {\"User-Agent\": \"github-file-history/1.0\"}\nNAME = re.compile(r\"^[A-Za-z0-9-]+/[A-Za-z0-9._-]+$\")\nNAME_COLS = (\"full_name\", \"fullName\", \"nameWithOwner\", \"repo_name\", \"repo\")\n\n\ndef opener(path):\n    return gzip.open(path, \"rt\", encoding=\"utf-8\", errors=\"replace\") if path.endswith(\".gz\") else open(path, encoding=\"utf-8\", errors=\"replace\")\n\n\ndef as_repo(s):\n    s = s.strip()\n    s = re.sub(r\"^(https?://)?(www\\.)?github\\.com/\", \"\", s)\n    s = re.sub(r\"\\.git$\", \"\", s)\n    parts = [p for p in s.split(\"/\") if p]\n    name = \"/\".join(parts[:2]) if len(parts) >= 2 and not s.startswith(\"#\") else None\n    return name if name and NAME.match(name) else None\n\n\ndef from_file(path):\n    with opener(path) as f:\n        return [r for r in (as_repo(l) for l in f) if r]\n\n\ndef from_search(path):\n    with opener(path) as f:\n        text = f.read()\n    try:\n        data = json.loads(text)\n    except ValueError:\n        start = min((text.find(c) for c in NAME_COLS if c in text), default=0)\n        start = text.rfind(\"\\n\", 0, start) + 1\n        rows = csv.DictReader(io.StringIO(text[start:]))\n        col = next(c for c in NAME_COLS if c in (rows.fieldnames or []))\n        return [r for r in (as_repo(row[col] or \"\") for row in rows) if r]\n    if isinstance(data, dict):\n        data = data.get(\"items\", data.get(\"repositories\", []))\n    out = []\n    for item in data:\n        name = item if isinstance(item, str) else next((item[c] for c in NAME_COLS if c in item), None)\n        if name is None and isinstance(item, dict) and \"owner\" in item and \"name\" in item:\n            owner = item[\"owner\"][\"login\"] if isinstance(item[\"owner\"], dict) else item[\"owner\"]\n            name = f\"{owner}/{item['name']}\"\n        r = as_repo(name or \"\")\n        if r:\n            out.append(r)\n    return out\n\n\ndef read_census(path):\n    \"\"\"Repo names from a census: a CSV with a repo_name column (the header may follow junk lines),\n    or one owner/repo per line. Names that are not owner/repo are dropped and counted.\"\"\"\n    with opener(path) as f:\n        lines = f.read().splitlines()\n    header = next((i for i, l in enumerate(lines[:100]) if \"repo_name\" in next(csv.reader([l]), [])), None)\n    if header is not None:\n        names = [(r.get(\"repo_name\") or \"\").strip() for r in csv.DictReader(io.StringIO(\"\\n\".join(lines[header:])))]\n    else:\n        names = [l.strip() for l in lines if l.strip() and not l.startswith(\"#\")]\n    good = [n for n in names if NAME.match(n)]\n    if len(good) < len(names):\n        bad = [n for n in names if not NAME.match(n)]\n        print(f\"census: dropped {len(bad)} lines that are not owner/repo, e.g. {bad[0][:60]!r}\", file=sys.stderr)\n    if not good:\n        sys.exit(f\"{path}: no owner/repo names found. Use a CSV with a repo_name column, or one owner/repo per line.\")\n    return good\n\n\ndef from_catalog(src, names, any_depth):\n    if src.startswith(\"http\"):\n        xml = urllib.request.urlopen(urllib.request.Request(src, headers=UA), timeout=60).read().decode(\"utf-8\")\n    else:\n        xml = open(src, encoding=\"utf-8\").read()\n    locs = re.findall(r\"https://agentlefs\\.com/context/([^<]+)\", xml)\n    total = xml.count(\"\")\n    out, seen = [], set()\n    for p in locs:\n        parts = p.split(\"/\")\n        if len(parts) < 3 or (not any_depth and len(parts) != 3):\n            continue\n        if names and parts[-1] not in names:\n            continue\n        if p not in seen:\n            seen.add(p)\n            out.append(p)\n    print(f\"sitemap urls {total}, files kept {len(out)}\", file=sys.stderr)\n    if total >= 20000:\n        print(\"note: the sitemap is at its 20,000-URL cap, so the lowest-starred files may be missing\", file=sys.stderr)\n    return out\n\n\ndef probe(repo, path, tries=4):\n    url = f\"https://raw.githubusercontent.com/{repo}/HEAD/{path}\"\n    code = None\n    for k in range(tries):\n        try:\n            with urllib.request.urlopen(urllib.request.Request(url, method=\"HEAD\", headers=UA), timeout=15) as r:\n                return r.status\n        except urllib.error.HTTPError as e:\n            if e.code == 404:\n                return 404\n            code = e.code\n        except Exception as e:  # timeouts, resets\n            code = type(e).__name__\n        time.sleep(2 ** k)\n    return code\n\n\ndef is_glob(p):\n    return any(c in p for c in \"*?[\")\n\n\ndef main():\n    ap = argparse.ArgumentParser(description=__doc__, formatter_class=argparse.RawDescriptionHelpFormatter)\n    ap.add_argument(\"source\", choices=[\"file\", \"search\", \"census\", \"catalog\"])\n    ap.add_argument(\"input\", nargs=\"?\", help=\"the file to read (catalog: optional local sitemap)\")\n    ap.add_argument(\"--path\", action=\"append\", default=[], help=\"path or glob to collect in every repo\")\n    ap.add_argument(\"--n\", type=int, help=\"census: repos to draw\")\n    ap.add_argument(\"--seed\", type=int, default=1)\n    ap.add_argument(\"--sampled\", help=\"census: also write the drawn repos here\")\n    ap.add_argument(\"--names\", nargs=\"+\", help=\"catalog: keep these file names\")\n    ap.add_argument(\"--any-depth\", action=\"store_true\", help=\"catalog: also files below the repo root\")\n    ap.add_argument(\"--probe\", action=\"store_true\")\n    ap.add_argument(\"--jobs\", type=int, default=16)\n    ap.add_argument(\"--limit\", type=int, default=0)\n    a = ap.parse_args()\n\n    if a.source == \"catalog\":\n        files = from_catalog(a.input or SITEMAP, a.names, a.any_depth)\n        if not a.path:\n            pairs = [(\"/\".join(p.split(\"/\")[:2]), \"/\".join(p.split(\"/\")[2:])) for p in files]\n            repos = None\n        else:\n            repos = list(dict.fromkeys(\"/\".join(p.split(\"/\")[:2]) for p in files))\n    else:\n        if not a.input:\n            sys.exit(f\"{a.source} needs an input file\")\n        if a.source == \"file\":\n            repos = from_file(a.input)\n        elif a.source == \"search\":\n            repos = from_search(a.input)\n        else:\n            if not a.n:\n                sys.exit(\"census needs --n\")\n            names = read_census(a.input)\n            repos = random.Random(a.seed).sample(names, min(a.n, len(names)))\n            print(f\"census {len(names)} repos, drew {len(repos)} (seed {a.seed})\", file=sys.stderr)\n            if a.sampled:\n                open(a.sampled, \"w\").write(\"\".join(r + \"\\n\" for r in repos))\n        repos = list(dict.fromkeys(repos)) if a.source != \"census\" else repos\n    if repos is not None:\n        if not a.path:\n            sys.exit(\"say which files to collect with --path\")\n        pairs = [(r, p) for r in repos for p in a.path]\n\n    if a.probe:\n        todo = [(r, p) for r, p in pairs if not is_glob(p)]\n        ok, errs = set(), 0\n        with cf.ThreadPoolExecutor(a.jobs) as ex:\n            for (r, p), code in zip(todo, ex.map(lambda j: probe(*j), todo)):\n                if code == 200:\n                    ok.add((r, p))\n                elif code != 404:\n                    print(f\"ERR {code} {r} {p}\", file=sys.stderr)\n                    errs += 1\n        print(f\"probed {len(todo)}, found {len(ok)}, errors {errs}\", file=sys.stderr)\n        pairs = [(r, p) for r, p in pairs if is_glob(p) or (r, p) in ok]\n    if a.limit:\n        pairs = pairs[: a.limit]\n    sys.stdout.write(\"\".join(f\"{r}/{p}\\n\" for r, p in pairs))\n\n\nif __name__ == \"__main__\":\n    main()\n","github-file-history/scripts/history.sh","#!/usr/bin/env bash\n# history.sh OWNER/REPO OUTDIR PATTERN [PATTERN ...]\n#\n# One bare, blob-less clone of a public repo, then the full history of every file that matches\n# a PATTERN. A pattern is a literal path (README.md, docs/intro.md) or a glob:\n#   *   any characters except /      ?  one character except /\n#   **  any characters including /   [abc]  one of a set\n# A glob is matched against every path that ever existed in the history, so files that were\n# deleted or renamed away are found too.\n#\n# For each file it writes OUTDIR/vers/___.tsv.gz, one row per\n# commit that touched the file, in git log order (newest first):\n#   sha  author_date (ISO 8601)  claude_coauthor (0/1)  base64(file body)\n# A commit that deleted the file has an empty body. Set HISTORY_FORMAT=tsv for plain .tsv files.\n#\n# It prints one manifest line per file written or already present: key  owner/repo  path.\n# Errors go to stderr: \"ERR clone \", \"ERR nohistory  \", or \"ERR blob \n#  \" when a version's content could not be fetched (that file is not written).\n# Each file's rows come from its own single-path git log, so a file's history is the same\n# whether you asked for it by name or through a glob.\nset -uo pipefail\nrepo=${1:?owner/repo}; out=${2:?output dir}; shift 2\n[ $# -gt 0 ] || { echo \"usage: history.sh OWNER/REPO OUTDIR PATTERN...\" >&2; exit 2; }\nmkdir -p \"$out/vers\"\nout=$(cd \"$out\" && pwd)\next=tsv.gz; [ \"${HISTORY_FORMAT:-gz}\" = tsv ] && ext=tsv\nkeyof() { local p=$1; echo \"${repo//\\//_}__${p//\\//_}\"; }\nisglob() { case \"$1\" in *[\\*\\?\\[]*) return 0;; *) return 1;; esac; }\n\n# Literal paths that are already collected need no clone.\nneed=0\nfor pat in \"$@\"; do\n  if isglob \"$pat\"; then need=1\n  elif [ -s \"$out/vers/$(keyof \"$pat\").$ext\" ]; then printf '%s\\t%s\\t%s\\n' \"$(keyof \"$pat\")\" \"$repo\" \"$pat\"\n  else need=1; fi\ndone\n[ $need = 1 ] || exit 0\n\nd=$(mktemp -d)\ntrap 'rm -rf \"$d\" 2>/dev/null' EXIT\nif ! timeout \"${CLONE_TIMEOUT:-300}\" git clone -q --bare --filter=blob:none \"https://github.com/$repo.git\" \"$d/g\" 2>/dev/null; then\n  echo \"ERR clone $repo\" >&2\n  exit 0\nfi\ncd \"$d/g\" || exit 0\n\n# glob -> anchored ERE\nglob_re() {\n  local g=$1 r=\"\" c i\n  for ((i = 0; i < ${#g}; i++)); do\n    c=${g:i:1}\n    if [ \"$c\" = \"*\" ] && [ \"${g:i+1:1}\" = \"*\" ]; then\n      if [ \"${g:i+2:1}\" = \"/\" ]; then r+=\"(.*/)?\"; i=$((i + 2)); else r+=\".*\"; i=$((i + 1)); fi\n    elif [ \"$c\" = \"*\" ]; then r+=\"[^/]*\"\n    elif [ \"$c\" = \"?\" ]; then r+=\"[^/]\"\n    elif [ \"$c\" = \"[\" ]; then r+=\"[\"\n    elif [ \"$c\" = \"]\" ]; then r+=\"]\"\n    elif [[ \"$c\" == [.^\\$+\\(\\)\\{\\}\\|\\\\] ]]; then r+=\"\\\\$c\"\n    else r+=\"$c\"; fi\n  done\n  echo \"^$r\\$\"\n}\n\npaths=()\nfor pat in \"$@\"; do\n  if isglob \"$pat\"; then\n    re=$(glob_re \"$pat\")\n    mapfile -t found < <(git log --format= --name-only --no-renames 2>/dev/null | grep -E \"$re\" | sort -u)\n    [ ${#found[@]} -gt 0 ] || { echo \"ERR nohistory $repo $pat\" >&2; continue; }\n    paths+=(\"${found[@]}\")\n  else\n    paths+=(\"$pat\")\n  fi\ndone\n\nzero=0000000000000000000000000000000000000000\n[ ${#paths[@]} -gt 0 ] || exit 0          # every pattern was a glob that matched nothing (already logged)\nprintf '%s\\n' \"${paths[@]}\" | awk '!seen[$0]++' > \"$d/paths.txt\"\n\n# Pass 1: the commits that touched each file, and the blob each one left.\nn=0\n: > \"$d/todo.tsv\"\nwhile IFS= read -r path <&3; do\n  key=$(keyof \"$path\")\n  dest=\"$out/vers/$key.$ext\"\n  if [ -s \"$dest\" ]; then printf '%s\\t%s\\t%s\\n' \"$key\" \"$repo\" \"$path\"; continue; fi\n  n=$((n + 1))\n  # One line per (commit, blob of PATH after the commit). --raw gives the new blob id.\n  # Default history simplification first (git log's usual view of one path). Only if that finds\n  # nothing, e.g. a file that lived only on a merged side branch, retry with --full-history.\n  for mode in \"\" \"--full-history\"; do\n    git log $mode --format='C%x09%H%x09%aI%x09%(trailers:key=Co-authored-by,valueonly,separator=;)' \\\n        --raw --no-abbrev --no-renames -- \"$path\" 2>/dev/null \\\n      | awk -F'\\t' -v f=\"$path\" '\n          /^C\\t/ { sha=$2; date=$3; co=(tolower($4) ~ /claude/) ? 1 : 0; next }\n          /^:/   { split($1, a, \" \"); if ($2 == f) print sha \"\\t\" date \"\\t\" co \"\\t\" a[4] }' > \"$d/c$n.tsv\"\n    [ -s \"$d/c$n.tsv\" ] && break\n  done\n  if [ ! -s \"$d/c$n.tsv\" ]; then\n    echo \"ERR nohistory $repo $path\" >&2\n    continue\n  fi\n  printf '%s\\t%s\\n' \"$n\" \"$path\" >> \"$d/todo.tsv\"\ndone 3< \"$d/paths.txt\"\n\n# Pass 2: fetch every needed blob in a few batches instead of one round trip per blob.\n# If the server refuses a batch, git falls back to fetching each blob when it is read.\ncat \"$d\"/c[0-9]*.tsv 2>/dev/null | cut -f4 | grep -v \"^$zero\\$\" | sort -u > \"$d/oids.txt\"\nif [ -s \"$d/oids.txt\" ]; then\n  xargs -n 500 git -c fetch.negotiationAlgorithm=noop fetch -q --no-tags --no-write-fetch-head \\\n      --recurse-submodules=no --filter=blob:none origin < \"$d/oids.txt\" >/dev/null 2>&1 || true\nfi\n\n# Pass 3: write each file's rows. A blob that cannot be read is an error, never a deletion:\n# the file is left out and logged as \"ERR blob\", so a re-run tries it again.\nwhile IFS=$'\\t' read -r i path <&3; do\n  key=$(keyof \"$path\")\n  dest=\"$out/vers/$key.$ext\"\n  ok=1\n  while IFS=$'\\t' read -r sha date co oid; do\n    if [ \"$oid\" = \"$zero\" ]; then b=\"\"\n    elif ! b=$(git cat-file blob \"$oid\" 2>/dev/null | base64 -w0; exit \"${PIPESTATUS[0]}\"); then\n      echo \"ERR blob $repo $path $sha\" >&2; ok=0; break\n    fi\n    printf '%s\\t%s\\t%s\\t%s\\n' \"$sha\" \"$date\" \"$co\" \"$b\"\n  done < \"$d/c$i.tsv\" > \"$d/rows.tsv\"\n  [ $ok = 1 ] || continue\n  if [ \"$ext\" = tsv ]; then mv \"$d/rows.tsv\" \"$dest\"; else gzip -n < \"$d/rows.tsv\" > \"$dest.tmp\" && mv \"$dest.tmp\" \"$dest\"; fi\n  printf '%s\\t%s\\t%s\\n' \"$key\" \"$repo\" \"$path\"\ndone 3< \"$d/todo.tsv\"\n","github-file-history/scripts/gharchive_census.py","#!/usr/bin/env python3\n\"\"\"Build a census of active public repos from GH Archive's free hourly files.\n\nGH Archive (https://www.gharchive.org) publishes every public GitHub event, one gzipped JSON file\nper hour, at https://data.gharchive.org/YYYY-MM-DD-H.json.gz. No account or token is needed.\nThis keeps the names of repos with at least one event of the chosen types in the chosen hours.\n\nUsage:\n  python3 gharchive_census.py --start 2026-10-01T00 --hours 24 [--every 1]\n                              [--types PushEvent] > census.txt\n\n  --start   first hour, UTC, as YYYY-MM-DDTHH\n  --hours   how many hours to read from --start\n  --every   read one hour in N (e.g. --hours 720 --every 24: one hour a day for 30 days)\n  --types   event types to count (default PushEvent: repos someone pushed to)\n\nOutput: one owner/repo per line, sorted, unique. make_list.py census reads it.\nOne hour is about 18 MB to download and gives about 40,000 repos (2026-10-01 12:00 UTC gave 39,397).\nA failed hour is reported on stderr and skipped.\n\"\"\"\nimport argparse\nimport gzip\nimport io\nimport json\nimport re\nimport sys\nimport urllib.request\nfrom datetime import datetime, timedelta\n\nNAME = re.compile(r\"^[A-Za-z0-9-]+/[A-Za-z0-9._-]+$\")\n\n\ndef main():\n    ap = argparse.ArgumentParser(description=__doc__, formatter_class=argparse.RawDescriptionHelpFormatter)\n    ap.add_argument(\"--start\", required=True)\n    ap.add_argument(\"--hours\", type=int, default=1)\n    ap.add_argument(\"--every\", type=int, default=1)\n    ap.add_argument(\"--types\", nargs=\"+\", default=[\"PushEvent\"])\n    a = ap.parse_args()\n    t0 = datetime.strptime(a.start, \"%Y-%m-%dT%H\")\n    types, repos, read = set(a.types), set(), 0\n    for h in range(0, a.hours, a.every):\n        t = t0 + timedelta(hours=h)\n        url = f\"https://data.gharchive.org/{t:%Y-%m-%d}-{t.hour}.json.gz\"\n        try:\n            raw = urllib.request.urlopen(urllib.request.Request(url, headers={\"User-Agent\": \"github-file-history/1.0\"}), timeout=120).read()\n        except Exception as e:  # missing hours exist in the archive\n            print(f\"ERR {url} {e}\", file=sys.stderr)\n            continue\n        with gzip.open(io.BytesIO(raw), \"rt\", encoding=\"utf-8\", errors=\"replace\") as f:\n            for line in f:\n                try:\n                    e = json.loads(line)\n                except ValueError:\n                    continue\n                if e.get(\"type\") in types:\n                    name = (e.get(\"repo\") or {}).get(\"name\", \"\")\n                    if NAME.match(name):\n                        repos.add(name)\n        read += 1\n        print(f\"{url}: {len(repos)} repos so far\", file=sys.stderr)\n    sys.stdout.write(\"\".join(r + \"\\n\" for r in sorted(repos)))\n    print(f\"read {read} hours, {len(repos)} repos\", file=sys.stderr)\n\n\nif __name__ == \"__main__\":\n    main()\n","github-file-history/scripts/collect.sh","#!/usr/bin/env bash\n# collect.sh LIST OUTDIR [JOBS]\n#\n# LIST: one line per wanted file as owner/repo/PATTERN, where PATTERN is a path or a glob\n# (README.md, package.json, .cursor/rules/*.mdc, docs/**/*.md). Blank lines and lines starting\n# with # are ignored. Paths with spaces are not supported. make_list.py writes this file.\n#\n# Lines are grouped by repo, so each repo is cloned once however many patterns it has.\n# history.sh runs JOBS repos at a time (default 8).\n# Writes OUTDIR/manifest.tsv (key, repo, path: one line per file collected) and OUTDIR/err.log.\n# Safe to re-run: finished files are skipped, and literal paths already on disk need no clone.\nset -uo pipefail\nlist=${1:?list of owner/repo/pattern}; out=${2:?output dir}; jobs=${3:-8}\nhere=$(cd \"$(dirname \"$0\")\" && pwd)\nmkdir -p \"$out/vers\"\nout=$(cd \"$out\" && pwd)\n: > \"$out/err.log\"\n\n# one line per repo: owner/repo pattern1 pattern2 ...\ngrep -v -E '^[[:space:]]*(#|$)' \"$list\" | awk '{\n    n = split($0, a, \"/\"); repo = a[1] \"/\" a[2]; pat = substr($0, length(repo) + 2)\n    if (!(repo in seen)) { seen[repo] = 1; order[++k] = repo }\n    pats[repo] = pats[repo] \" \" pat\n  } END { for (i = 1; i <= k; i++) print order[i] pats[order[i]] }' > \"$out/.repos\"\n\nexport HISTORY_OUT=\"$out\" HISTORY_SH=\"$here/history.sh\"\n# xargs -L1 passes each repo line as arguments; history.sh prints manifest lines on stdout.\nxargs -P \"$jobs\" -L 1 bash -c 'bash \"$HISTORY_SH\" \"$0\" \"$HISTORY_OUT\" \"$@\"' < \"$out/.repos\" \\\n  > \"$out/.manifest\" 2>>\"$out/err.log\"\nsort -u \"$out/.manifest\" > \"$out/manifest.tsv\"\nrm -f \"$out/.manifest\" \"$out/.repos\"\n\nlisted=$(grep -c -v -E '^[[:space:]]*(#|$)' \"$list\")\necho \"listed $listed  files collected $(wc -l < \"$out/manifest.tsv\")  errors $(grep -c \"^ERR\" \"$out/err.log\")\"\n","github-file-history/examples/smoke.txt","# smoke test: two README histories, one glob, and one repo that does not exist\nchalk/chalk/readme.md\njshttp/mime-types/*.md\nnobody-here/not-a-repo-xyz/README.md\n","github-file-history/examples/readme-package/repos.txt","# five public repos for the README / package.json example\nchalk/chalk\nexpressjs/cors\nsindresorhus/ky\npallets/itsdangerous\njshttp/mime-types\n","github-file-history/examples/readme-package/list.txt","chalk/chalk/package.json\nchalk/chalk/[Rr][Ee][Aa][Dd][Mm][Ee]*\nexpressjs/cors/package.json\nexpressjs/cors/[Rr][Ee][Aa][Dd][Mm][Ee]*\nsindresorhus/ky/package.json\nsindresorhus/ky/[Rr][Ee][Aa][Dd][Mm][Ee]*\npallets/itsdangerous/package.json\npallets/itsdangerous/[Rr][Ee][Aa][Dd][Mm][Ee]*\njshttp/mime-types/package.json\njshttp/mime-types/[Rr][Ee][Aa][Dd][Mm][Ee]*\n","github-file-history/SKILL.md","# GitHub file history\n\nPulls the full git history of chosen files from many public GitHub repos, one TSV per file.\n\nEach file becomes one TSV (gzipped by default) with a row per commit that touched it, plus a\nmanifest. The repo list can come from a plain file, a GitHub search export, a random draw from a\ncensus list, or the agentleFS public catalog. It needs no GitHub token: it uses public\n`git clone` and `raw.githubusercontent.com` only, and neither counts against the REST API rate\nlimit.\n\nExample output (the smoke test below, 2026-10-11; base64 bodies cut short):\n\n```\n$ bash $SKILL/scripts/collect.sh $SKILL/examples/smoke.txt /tmp/gfh-smoke 4\nlisted 3  files collected 4  errors 1\n$ zcat /tmp/gfh-smoke/vers/jshttp_mime-types__README.md.tsv.gz | head -2\n80cdcb346c66fb39ab17be88a0a1031ffaee678f\t2026-01-08T15:03:44-06:00\t0\tIyBtaW1lLXR5cGVzCgpbIVtOUE0g...\n3aabd6896010e1d2c0c75d5cea36f90cd403b623\t2025-11-25T11:12:26+01:00\t0\tIyBtaW1lLXR5cGVzCgpbIVtOUE0g...\n```\n\nUse it when:\n- you want to see how a kind of file changes over time across projects: READMEs, package.json,\n  Dockerfiles, CI workflows, agent instruction files, editor rules\n- you need input for `text-rule-trends`\n\n## Requirements\n\n- `bash`, `git` 2.22+ (partial clone and `%(trailers)`), `gzip`, `base64`, `awk`, `timeout`\n- Python 3.8+, standard library only (for `make_list.py`)\n- Network access to `github.com`, and to `raw.githubusercontent.com` for `--probe`.\n  The `catalog` source also reads `agentlefs.com`.\n\n## Inputs\n\nA list file with one wanted file per line, as `owner/repo/PATTERN`. A pattern is a path or a glob:\n\n| pattern | matches |\n|---|---|\n| `README.md` | that path only |\n| `*.md` | any `.md` file at the root |\n| `.cursor/rules/*.mdc` | `.mdc` files in that folder (not below it) |\n| `docs/**/*.md` | `.md` files anywhere under `docs/` |\n| `[Rr][Ee][Aa][Dd][Mm][Ee]*` | README, readme.md, README.rst, ... |\n\nA glob is matched against every path that ever existed on the default branch. So it also finds\nfiles that were later deleted or renamed. Each match becomes its own file history.\n\nYou can write the list by hand, or build it with `make_list.py` from one of four sources.\n\n## Steps\n\nRun the scripts by their path and write outputs to a folder of your own, never inside the skill\nfolder. Below, `SKILL` is this skill's folder.\n\n### 1. Make the list\n\n```bash\n# a plain list of repos (owner/repo or GitHub URLs, one per line)\npython3 $SKILL/scripts/make_list.py file repos.txt --path README.md --path package.json > list.txt\n\n# a GitHub search export: gh search repos ... --json fullName, a saved /search/repositories\n# response, or a CSV with full_name / repo_name / nameWithOwner\npython3 $SKILL/scripts/make_list.py search repos.json --path '.cursor/rules/*.mdc' > list.txt\n\n# a random draw from any census list, keeping only files that exist today\npython3 $SKILL/scripts/make_list.py census census.csv.gz --n 1000 --seed 1 \\\n    --path Dockerfile --probe --sampled drawn.txt > list.txt\n\n# the agentleFS public catalog: its own agent files, or its repos crossed with --path\npython3 $SKILL/scripts/make_list.py catalog --names CLAUDE.md AGENTS.md > list.txt\npython3 $SKILL/scripts/make_list.py catalog --path package.json --limit 50 > list.txt\n```\n\n- `--probe` sends a HEAD request to `https://raw.githubusercontent.com//HEAD/` and\n  keeps the pairs that answer 200. It works for literal paths only. Globs pass through.\n  Non-404 errors are retried 4 times, then printed as `ERR   `.\n  `--jobs` (default 16) sets how many probes run at once.\n- `census` reads a CSV with a `repo_name` column (gzipped or not; the header may come after junk\n  lines) or one `owner/repo` per line. Lines that are not a valid `owner/repo` are dropped and\n  counted on stderr. It draws with `random.Random(seed).sample`. `--sampled` keeps the drawn\n  repos, so you have the denominator on record.\n- **No census ships with this skill.** Make one from GH Archive's free hourly event files:\n  ```bash\n  python3 $SKILL/scripts/gharchive_census.py --start 2026-10-01T00 --hours 720 --every 24 > census.txt\n  ```\n  That reads one hour a day for 30 days and keeps every repo with a push. One hour is about\n  18 MB and gave 39,397 repos (2026-10-01 12:00 UTC, tested 2026-10-11). Any list of public\n  repos works instead, in either format above. Your list decides what \"random\" means, so write\n  down how you made it.\n- `catalog` reads `https://agentlefs.com/context/sitemap.xml`. It lists agent instruction files\n  in public repos with 1,000+ stars. Without `--path`, it lists the catalog's files (root only\n  unless `--any-depth`, filtered by `--names`). With `--path`, it is a source of popular repos.\n\n### 2. Collect histories\n\n```bash\nbash $SKILL/scripts/collect.sh list.txt data/mysample 8\n```\n\n`collect.sh` groups the list by repo, so each repo is cloned once. For each repo, `history.sh`\nmakes a bare, blob-less clone (`git clone --bare --filter=blob:none`). It expands any globs\nagainst the history, lists the commits that touched each file, and fetches only those blobs, in\nbatches of 500 per `git fetch` rather than one round trip each. If a blob still cannot be read,\nthat file is not written and `ERR blob` is logged, so a failed fetch never looks like a deletion,\nand a re-run tries the file again.\nIt is safe to re-run. Finished files are skipped, and literal paths already on disk skip the clone.\n\nOne repo on its own: `bash $SKILL/scripts/history.sh owner/repo OUTDIR README.md 'docs/*.md'`.\n\nSettings, as environment variables:\n- `CLONE_TIMEOUT` (seconds, default 300) limits each clone. Huge repos can need more.\n- `HISTORY_FORMAT=tsv` writes plain `.tsv` files instead of `.tsv.gz`, for stores that hold text only.\n\n## Outputs\n\n```\ndata//\n  manifest.tsv            key  owner/repo  path   (one line per file collected)\n  err.log                 always written; \"ERR clone \", \"ERR nohistory  \", or\n                          \"ERR blob   \" when a version could not be fetched\n  vers/.tsv.gz       one per file; key = ___\n```\n\nEach `vers` row is one commit that touched the file, in `git log` order, with four tab-separated\nfields. That order is newest first by commit graph, not strictly by author date: author dates can\ngo backwards (rebases, cherry-picks, wrong clocks, mixed UTC offsets). Sort by the parsed date\nif you need date order; `text-rule-trends` does.\n\n| field | meaning |\n|---|---|\n| sha | commit id |\n| author_date | ISO 8601 author date with offset, e.g. `2026-03-31T14:14:16-07:00` |\n| claude_coauthor | 1 if a `Co-authored-by` trailer mentions \"claude\", else 0 |\n| body | the file's content after that commit, base64 (no newlines). Empty if the commit deleted it |\n\n`text-rule-trends` reads this format, `.tsv.gz` or `.tsv`.\n\n## Examples\n\n### README and package.json in five repos (`examples/readme-package/`)\n\n`repos.txt` names five public repos. `list.txt` was made from it:\n\n```bash\npython3 $SKILL/scripts/make_list.py file $SKILL/examples/readme-package/repos.txt \\\n    --path package.json --path '[Rr][Ee][Aa][Dd][Mm][Ee]*' > list.txt\nbash $SKILL/scripts/collect.sh list.txt /tmp/readme-package 5\n# listed 10  files collected 12  errors 1\n```\n\nOn 2026-10-11 this took 3.4 seconds. The README glob found 7 files in 5 repos, including\nrenames: `expressjs/cors` had `README.markdown` and then `README.md`, and `pallets/itsdangerous`\nhad `README`, then `README.rst`, then `README.md`. The one error is `pallets/itsdangerous`, a\nPython project with no `package.json`. Row counts grow as the repos get new commits.\n\n### Smoke test (`examples/smoke.txt`)\n\nTwo README histories, one glob, and one repo that does not exist. A few seconds:\n\n```bash\nbash $SKILL/scripts/collect.sh $SKILL/examples/smoke.txt /tmp/gfh-smoke 4\n# listed 3  files collected 4  errors 1\n```\n\nThe glob `jshttp/mime-types/*.md` found `HISTORY.md`, `README.md` and `SOURCES.md` (2026-10-11).\nThe error is `ERR clone nobody-here/not-a-repo-xyz`. Every `vers` row should have 4 fields, and\neach body should decode with `base64 -d`.\n\n**Worked example.** These skills were first written for the agentleFS article \"CLAUDE.md best\npractices\" (https://agentlefs.com/blog/claude-md-best-practices), which studied 5,786 CLAUDE.md and\nAGENTS.md files. Its data, settings and step-by-step recipe live outside the skills, in the\n`article-claude-md-best-practices` folder (`RECIPE.md` and `data/`).\n\n## Limits and caveats\n\n- **Your repo list decides what your sample means.** A random draw from a census is only as\n  random as the census. The catalog covers popular repos (1,000+ stars, licences that allow\n  redistribution), which are not typical repos. Write down where the list came from.\n- **The catalog sitemap has a cap of 20,000 URLs**, most stars first. It was at the cap on\n  2026-10-11. The script prints a note when the cap is reached. The catalog changes every week.\n- **`--probe` only sees files that exist today on the default branch.** Deleted or renamed files\n  are missed. Globs are not probed, so they find deleted files too, but they cost a clone per repo.\n- **History follows one path, with git's default history simplification.** A file that only ever lived on a side branch that was merged has no history\n  that way; for those (and only those) `history.sh` retries with `--full-history`, so they are\n  collected instead of logged as nohistory. Histories that the default finds are never changed.\n  A rename looks like a deletion plus a new file. A glob that matches\n  both names gives you both files. Dates are author dates, which people can set to anything.\n- **Keys can collide** in rare cases, because `/` and `_` both become `_` (`a/b_c` and `a_b/c`).\n  Paths with spaces are not supported.\n- **Rate limits.** Clones and raw HEAD requests need no token, but GitHub throttles heavy use.\n  Keep `--jobs` and the collect parallelism modest (8 to 24). `api.github.com` is not used. If you\n  add API calls of your own, read the token from `GITHUB_TOKEN` and never print it.\n- **Speed.** Blobs are fetched in batches, so the README example (12 files, about 1,200 versions)\n  took 3.4 seconds on 2026-10-11. Blob-less clones still download every commit and tree, so huge\n  monorepos can take minutes each.\n- **Licences.** Collecting public files to study them is one thing. Redistributing their text is\n  another. Only republish file contents (including these TSVs) for repos whose licence allows it.\n\n## Credits\n\n- **GH Archive** (https://www.gharchive.org), by Ilya Grigorik: its free hourly files of public\n  GitHub events are what `gharchive_census.py` reads to build a census of active repos. Nothing\n  from it ships with this skill. Its front page states no licence for the data (checked\n  2026-10-11); the events are GitHub's public timeline, under GitHub's terms.\n- **git partial clone** (`git clone --filter=blob:none`, git 2.22+): blob-less clones, with only\n  the needed blobs fetched in batches, are what make many repos cheap. git (GPL-2.0) is called as\n  an installed program, not bundled.\n- File contents collected belong to their repos' authors, under each repo's licence (see Limits).\n\n","github-file-history/LICENSE","clear-writing/SKILL.md","# Clear writing\n\nRewrite prose so a smart reader who is new to the subject understands it on one read. Write it the way you would explain it to a friend sitting next to you: in order, out loud, and all the way to the fix. Keep the author's voice and every fact.\n\nShort is not the goal. Understood is the goal. A sentence that is tight but leaves the reader asking \"so what do I do?\" is not finished.\n\n## The worked example\n\nThe original:\n\n> Claude Code subagents and agent teams start with your MCP tools and the lead's permission mode, so a researcher meant only to read a folder can also rewrite it.\n\nIt assumes the reader knows \"the lead\", \"permission mode\" and \"MCP tools\". It packs three ideas into one sentence. And it stops at the problem.\n\nA clipped rewrite (better, but not the target):\n\n> When Claude Code starts a helper agent, the helper gets every tool and every permission you have. So an agent you asked only to read a folder can also rewrite it.\n\nThis fixes the jargon and the order. But it reads like a warning label. The reader learns what goes wrong and is left there, with no idea what to do about it.\n\nThe target, in the author's own words:\n\n> When Claude Code starts a helper agent, the helper gets every tool and every permission you have. So what could happen is that this agent you asked for could also happen to rewrite it. The better way is to give each agent its own agentleFS identity, limited to one folder, and then any mistake stays inside that folder.\n\nWhat it does that the clipped one does not:\n\n- It talks. \"So what could happen is…\" and \"The better way is…\" are how a person explains something out loud. The reader is being walked through it, not handed a summary.\n- It follows the whole chain: cause (the helper gets everything), what goes wrong (it can rewrite what you only asked it to read), the fix (one identity, one folder), and why the fix works (a mistake stays in that folder).\n- It ends on the payoff, not the problem.\n- It is a little loose: \"could also happen to\", \"this agent you asked for\". That looseness is the voice. Leave it.\n\n## Rules\n\n1. **Talk the reader through it.** Write the way you would say it to a friend. Connecting phrases that carry the reader from one step to the next are welcome: \"So what could happen is…\", \"The better way is…\", \"and then…\", \"which means…\". They are not filler; they are the path.\n2. **Follow the chain all the way.** Cause, then what goes wrong, then the fix, then why the fix works. Never stop at the problem. If the piece names a risk, it also names what to do and what that buys you.\n3. **End on the payoff.** The last thing the reader takes away is what gets better, not what can go wrong.\n4. **Flow over compression.** Trade a few words for a sentence the reader does not have to re-read. Do not merge, cut or clip a sentence just to make it shorter. Split one only when it carries two ideas the reader would have to untangle.\n5. **Plain words.** Write for someone who has used a chatbot but never read the docs. Prefer \"use\" to \"leverage\", \"about\" to \"approximately\", \"can read\" to \"has read access\", \"folder\" to \"scope\".\n6. **No undefined jargon.** Explain a term a newcomer would not know in one short clause the first time, or replace it. For a developer audience, keep standard developer terms but still define the product's own.\n7. **Name who does what.** \"The server checks the token\", not \"the token is checked\".\n8. **Concrete over abstract.** Prefer an example (\"an agent asked to summarise `plans/` can delete `plans/`\") to a category (\"excessive privilege\").\n9. **Cut empty filler, keep spoken connectors.** Delete words that carry nothing: \"basically\", \"simply\", \"really\", \"in order to\", \"it's worth noting\", \"note that\", throat-clearing, and a sentence that repeats the one before it. Keep the words that move the reader to the next step, even when they sound casual.\n10. **Keep the voice, loose edges included.** Keep first person, the author's examples, jokes, opinions and their natural, slightly loose spoken phrasing. Do not polish it into textbook prose. Fix a phrase only when the reader genuinely cannot tell what it means.\n\n## Do not change\n\n- Facts, numbers, dates, names, claims and recommendations. If you cannot tell what a sentence claims, leave it and flag it.\n- Links and URLs, images and their alt text's meaning, code blocks, commands, config, and the facts in tables (cell wording may be tightened).\n- Frontmatter. Headings' meaning (wording may be shortened).\n- Notes in [square brackets] that read like the author's notes to themselves.\n- The author's spoken phrasing, when it is clear enough to follow.\n\n## How to work\n\n1. Read the whole piece first. Write down its point in one sentence, and the fix it recommends. If it names a problem and no fix, flag that.\n2. Edit paragraph by paragraph: put the point where the reader needs it, then lay out cause, what goes wrong, fix and payoff in that order. Then plain words, jargon, filler.\n3. Read it back out loud as the newcomer. It should sound like someone talking. Any sentence you had to read twice gets another pass; any sentence that sounds like a label gets loosened.\n4. Check the \"Do not change\" list against the original: same links, images, code and facts.\n5. Report the worst sentence before and after, any problem left without a fix, and any sentence you left because its claim was unclear.\n\n## Checklist\n\n- [ ] It reads like someone explaining it out loud.\n- [ ] Every problem is followed by its fix and why the fix works.\n- [ ] The reader ends on the payoff.\n- [ ] No sentence the reader has to read twice.\n- [ ] Every term a newcomer would stumble on is explained once or gone.\n- [ ] Nobody hidden behind the passive voice.\n- [ ] Empty filler gone; spoken connectors kept.\n- [ ] Links, images, code, numbers and frontmatter unchanged.\n- [ ] It still sounds like the author, loose edges and all.\n","clear-writing/LICENSE","clerkState",{"_250":251},"__internal_clerk_state",{"_252":253,"_254":255,"_256":32,"_257":32,"_258":20,"_259":32,"_260":32,"_261":-7,"_262":-7,"_263":-7,"_264":-7,"_265":266,"_267":-7,"_268":-7,"_269":-7,"_270":-7,"_271":-7,"_272":20,"_273":20,"_274":20},"__clerk_ssr_state",{"_284":285,"_286":-5,"_287":-5,"_288":-5,"_289":-5,"_290":-5,"_291":-5,"_292":-5,"_293":-5,"_294":-5,"_295":-5,"_296":297,"_298":299,"_300":301,"_276":20},"__publishableKey","pk_live_Y2xlcmsuYWdlbnRsZWZzLmNvbSQ","__proxyUrl","__domain","__isSatellite","__signInUrl","__signUpUrl","__signInForceRedirectUrl","__signUpForceRedirectUrl","__signInFallbackRedirectUrl","__signUpFallbackRedirectUrl","__clerk_debug",{"_275":20,"_276":20,"_277":32,"_278":279,"_280":32,"_281":255,"_282":20,"_283":32},"__clerkJSUrl","__clerkJSVersion","__clerkUIUrl","__clerkUIVersion","__prefetchUI","__telemetryDisabled","__telemetryDebug","__unsafeDisableDevelopmentModeConsoleWarning","isSignedIn","isAuthenticated","proxyUrl","reason","session-token-and-uat-missing","message","publishableKey","isSatellite","domain","tokenType","session_token","sessionClaims","sessionId","sessionStatus","userId","actor","orgId","orgRole","orgSlug","orgPermissions","factorVerificationAge","getToken",["SingleFetchFallback"],"has",["SingleFetchFallback"],"debug",["SingleFetchFallback"]]