cracktro
Skill5.6k starsChanged 6 months agoMITView on GitHub
Edit, preview, and verify the standalone interactive talk deck in docs/public/cracktro (slides, SVG diagrams, transitions, WebAudio sound).
Trending
Quiet since March.thin dataabout the whole repository
Who's talking about github/gh-aw
- A Hacker News story“GitHub Agentic Workflows”Hacker News · Feb 11, 2026 · 2 points
- @kcieslak.io“We see that a lot in our OSSed AW repository which uses fully agentic development - github.com/github/gh-aw …”Bluesky · Mar 24, 2026 · 1 likes
- @github-trending.bsky.social“🚀 Skyrocketing! 🚀 (200+ new stars) 📦 github / gh-aw ⭐ 2,482 (+213) 🗒 Go GitHub Agentic Workflows”Bluesky · Feb 15, 2026
The biggest 3 of 3 mentions linking to the repository that our crawl found. Each line is their own headline.
In this repository115 other files it gives its agents.
- AGENTS.mdAGENTS.md
- docs/src/content/docs/engines/claude.mdCLAUDE.md
- docs/src/content/docs/reference/engines/claude.mdCLAUDE.md
- pkg/workflow/data/engines/claude.mdCLAUDE.md
- .squad/templates/copilot-instructions.mdCopilot instructions
- playwright-cliSkill
- agentic-workflowsSkill
- awf-release-integratorSkill
- checkout-credential-reviewSkill
- console-renderingSkill
- copilot-reviewSkill
- create-canvasSkill
- create-connectorSkill
- custom-agentsSkill
- custom-engine-implementationSkill
- debugging-workflowsSkill
- developer-code-organizationSkill
- developer-internalsSkill
- developer-releaseSkill
- developer-securitySkill
- developerSkill
- docs-designSkill
- documentationSkill
- error-messagesSkill
- error-pattern-safetySkill
- error-recovery-patternsSkill
- gh-agent-sessionSkill
- gh-agent-taskSkill
- github-copilot-agent-tips-and-tricksSkill
- github-discussion-querySkill
- github-issue-querySkill
- github-labels-querySkill
- github-mcp-serverSkill
- github-pr-querySkill
- github-scriptSkill
- github-workflows-querySkill
- go-codemodSkill
- go-lintersSkill
- http-mcp-headersSkill
- issue-markdownSkill
- javascript-refactoringSkill
- jqschemaSkill
- messagesSkill
- model-routing-goldenSkill
- npmSkill
- open-pull-requestSkill
- operational-value-designerSkill
- optimize-agentic-workflowSkill
- otel-queriesSkill
- pr-finisherSkill
- prompt-token-efficiencySkill
- pr-to-go-linterSkill
- reportingSkill
- resolve-merge-conflictsSkill
- restricted-tool-triageSkill
- review-agentic-workflowsSkill
- security-advisory-compat-enforcerSkill
- sergo-examplesSkill
- skillz-integrationSkill
- ssl-skill-normalizerSkill
- The first 60 shown.
The file
We split this skill into the rules it gives an agent
4 rules. About 0 are common, 1 is distinctive, and 3 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
3
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 8
Use Claude Opus 5.5 (claude-opus-5.5) for deck work, including any subagents you delegate to.
- Very specific: line 63
Never use a real GitHub user's avatar, handle or name for a person, in slides, speaker notes or comments.
- Distinctive: line 65
Follow https://brand.github.com/graphic-elements/mascots: use mascots sparingly, never as logos, and never on serious topics (security, …
- Very specific: line 89
Run from docs/ after npm ci: npx playwright test tests/talk.spec.ts.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Cracktro talk deck
Use Claude Opus 5.5 (claude-opus-5.5) for deck work, including any subagents you delegate to.
The deck is published at https://github.github.com/gh-aw/cracktro/. It is a deck laid out as horizontal bands (Prezi-style) with no dependencies, separate from the docs homepage slideshow.
Files
docs/public/cracktro/index.html: loadsdeck.jsand holds the SEO/Open Graph/Twitter meta (absolutehttps://github.github.com/gh-aw/cracktro/URLs).docs/public/cracktro/og-image.png: 1200x630 social card rendered from slide 1 with the dots/count/sound/progress UI hidden. Keep it outsideassets/(the asset test only allows avatars there).docs/public/cracktro/deck.js: about 575 KB holding everything (CSS, slides, navigation, synth, background). Never read it whole. Usegrep -non the// ---------- <section> ----------and// <Slide title>comment markers, thenview_range.docs/public/cracktro/assets/avatars/{mona,ducky}.jpg: the fake users Mona and Ducky (GitHub's mascots), plusmona-full.png(transparent full-body Mona for the Trust the process slide); the only allowed assets, which a test enforces. Diagrams are inline SVG built indeck.js(for example slide 2's CI-to-agenticGA_RUNSpipeline).docs/public/cracktro/fonts/: Monaspace Neon plus its OFL licence. Mona Sans comes fromdocs/public/fonts/.docs/tests/talk.spec.ts: Playwright tests. They read.github/skills/cracktro/slides.md..github/skills/cracktro/slides.md: the single ordered slide list (headings, band headers included, and which notes must say illustrative). A slide's number is its row; never write slide numbers or counts anywhere else. Edit one row per slide in the same change asdeck.js.docs/README.md, "Interactive talk" section: user-facing description (keys, public-content rules).
deck.js layout (search for these markers)
| Marker | Contents |
---------- styles ---------- | A single CSS template string. Slide-specific classes use prefixes such as fm-, atk-, rpa-, mra-, eph-, fv-, sec-, ch-, pf- and bh- (band header slides). Check that a new prefix is unused before adding it. |
---------- helpers ---------- | h(tag, attrs, ...kids) creates DOM; letters(text, cls) builds per-letter headings with classes like l-grad and l-out; resetIndex(). |
---------- slides ---------- | One { ... slides.push({...}) } block per slide, each preceded by a // Title: summary comment. |
const OCTI | Octicon path map (calendar, issue, pr, merge, person, copilot, ...) used by ico(name,x,y,size,fill). |
---------- sound: | WebAudio synth sfx: slide, band, step, cut, techno, pop, key, compile, review, notify, omen, route, mr, toggle, unlock. MUSIC holds one theme per band; slide(dir, music, shift) transposes tone/bell/pad by shift semitones. |
---------- band themes ---------- | THEME: per-band background, grid, angle, accent, red wash and music (see mechanics.md). |
---------- navigation ---------- | go(), show(n, horiz), keys, dots, notes, hash. |
---------- animated background ---------- | Canvas background and grid driven by the current band's THEME. |
The stage is a fixed 1920x1080 frame (W/H), scaled to fit the window.
Slide object
slides.push({
notes: "...", // speaker notes (N key); public text
group: "security", // band id; picks the THEME (colours, grid, music, red wash); crossing groups slides up
techno: 1, // optional techno layer level
el: h("section", { class: "slide" }, ...),
enter(dir) {}, // reset per-slide state; dir = 1 forward, -1 back
tick(t) {}, // called every frame, ONLY for the current slide
onNext() {}, onPrev() {}, onKey(e) {}, // return true to consume (in-slide steps)
});Reference files (load only what the task needs)
| File | Load when |
mechanics.md | Changing bands, slide order or group, navigation, transitions, backgrounds, audio, or animated SVG flows (EDGES/SEQ/LANE). |
content-rules.md | Adding any public text, data, avatar or image; contains the per-slide Exception bullets for measured data. |
workflow-prompts.md | Adding or refreshing a slide's /agentic-workflows prompt line (PROMPTS table and phone dialog); holds the prompt-mining algorithm and the exact PR/task/session references. |
data-refresh.md | Adding, changing or refreshing a data slide; holds the refresh mini-skill convention and the slide→refresh/<slug>.md index. |
People and mascots
- Never use a real GitHub user's avatar, handle or name for a person, in slides, speaker notes or comments. Never fetch
github.com/<login>.png. - People are only the fake users
monaandducky, drawn fromassets/avatars/{mona,ducky}.jpg. Mona is the maintainer; Ducky is the external contributor (community user). ADevelopercan be either: mix them across slides, and use both when a slide shows two developers. - Follow https://brand.github.com/graphic-elements/mascots: use mascots sparingly, never as logos, and never on serious topics (security, enterprise, money). Use role icons there.
- Never generate or redraw mascots. Both avatars are 160x160 clips of official brand.github.com art in the modern three-quarter 3D style:
mona.jpgis the dark-mode purple Mona frommona-2026-versions.png,ducky.jpgthe middle Ducky frommascots-03.png. Their backgrounds are recoloured to the deck's#0d1117(never white) so they blend into the dark avatar circles.
Core rules (details in content-rules.md): everything is public; no real GitHub avatars, handles or people, only the fake users Mona and Ducky (per the brand mascot guidelines); no 40-hex SHAs or private repos; the only URLs are https://github.github.com/gh-aw/introduction/overview/ and the For PMs source link https://github.com/chrizbo/agentics-beyond-code; speaker notes of illustrative slides must match /illustrative|synthetic|simulated/i.
Preview loop
There is no build step. Serve the static folder and open it in the browser canvas:
python3 -m http.server 4329 --bind 127.0.0.1 --directory docs/public # run async + detached
# open http://127.0.0.1:4329/cracktro/#11 (#N = row N of slides.md)After each edit:
- Run
node --check docs/public/cracktro/deck.js. - In the canvas, run
location.reload(), wait, and take a screenshot. - To jump to a slide, click
[aria-label="Slide N"]or set the hash. - Dispatch key events on
window, for exampledispatchEvent(new KeyboardEvent('keydown',{key:'ArrowRight'})). - To catch a transition, read
getComputedStyle(el).transformabout 200 ms after navigating.
Tests
Run from docs/ after npm ci: npx playwright test tests/talk.spec.ts. The tests use the Astro dev server on localhost:4321 at the base path /gh-aw/cracktro/. If node_modules is missing, say the tests were not run rather than skipping silently.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

