nutrition-mcp-setup
Skill4 starsChanged 4 months agoMITView on GitHub
Stand up the `nutrition` MCP for the "Nutrition & Chef" add-on on a new machine and wire it into both Claude clients — the simplest of the optional add-ons (a pure thin fetch-wrapper over the board's /api/nutrition/* routes, like the calendar server; NO sidecar, NO external repo, all in-repo at mcp/nutrition-server). It is exposed to Claude Code via a supergateway + launchd BRIDGE on $NUTRITION_BRIDGE_PORT/:8007 and to Claude Cowork Desktop as a direct stdio command, and the add-on must be ENABLED (Settings.addons.nutrition.enabled) for its tools' WRITES to land + its /nutrition nav to appear. Use when setting up the nutrition/chef add-on on a new machine, when Cowork or Code can't see the `nutrition` server, when the nutrition bridge (:8007) is down, when log_food / plan_meal / add_pantry_item write but the board 404s them, or when enabling/disabling the Nutrition & Chef add-on.
- Reads credentials
- Installs packages
- Sends data out
In this repository17 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- mcp/CLAUDE.mdCLAUDE.md
- add-an-addonSkill
- backup-recoverySkill
- body-mcp-setupSkill
- cos-setupSkill
- cos-upgradeSkill
- debug-cowork-mcp-issuesSkill
- fitness-mcp-setupSkill
- guard-setupSkill
- hub-handoverSkill
- mcp-bridge-setupSkill
- openwhispr-mcp-setupSkill
- setup-vaultSkill
- spoke-setupSkill
- voice-operationsSkill
- whatsapp-mcp-setupSkill
The file
We split this skill into the rules it gives an agent
5 rules. About 0 are common, 0 are distinctive, and 5 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈0
Distinctive · estimate
This author's own choices. Worth a look.
5
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 64
Run the loader preamble as the first line of every shell block below — it exports $REPO_ROOT, $NODE_BIN, $BREW_PREFIX, $SUPERGATEWAY_BIN, …
- Very specific: line 213
Delete the smoke-test entry afterwards (delete_food_log with the FOOD-id from (a)), or leave it — it's harmless.
- Very specific: line 299
Not found." is almost always *the add-on is off* (§6), not a bridge fault — check /api/addons.
- Very specific: line 312
The board must be reachable on
CRM_BASE_URL: Every tool is a fetch to /api/nutrition/*. - Very specific: line 313
If the board app isn't up on :3000 (or wherever CRM_BASE_URL points), tool calls fail with a connection error — start the board (cd board …
≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Nutrition & Chef MCP setup (the simplest add-on — thin fetch-wrapper bridge :8007)
Why this exists / architecture
nutrition is the first built-in add-on ("Nutrition & Chef" — food log + pantry + meal
plan). Unlike the openwhispr / whatsapp add-ons it has no external repo and no sidecar: the
server lives in this Cos checkout at mcp/nutrition-server/server.mjs, and it is the simplest
archetype in the whole system — a thin fetch wrapper over the board's /api/nutrition/*
HTTP routes on CRM_BASE_URL, exactly like the calendar server. It makes no LLM calls and
needs no secret; the intelligence (estimating calories/macros) lives in the operator skill
(/nutrition-chef), not here — the MCP just stores numbers.
It is a single Node stdio process and registers into Cos the same two ways the core servers do:
- Claude Code reaches it over HTTP via a supergateway + launchd BRIDGE on
:8007($NUTRITION_BRIDGE_PORT, labelcom.chiefofstaff.mcp-nutrition, IN.mcp.json). - Claude Cowork Desktop spawns it as a direct stdio
commandentry inclaude_desktop_config.json— identical to how mcp-bridge-setup wires board/calendar/etc.
Claude Code ──HTTP──> localhost:8007/mcp ──supergateway(launchd)──> node nutrition-server (stdio)
│ fetch (actor: agent)
└──HTTP──> board /api/nutrition/* (:3000)
Cowork Desktop ──spawns stdio directly──────────────────────────────────> node nutrition-server (§4)This is an ADD-ON — its WRITES are GATED, its READS are open. The "Nutrition & Chef" add-on must be ENABLED (
Settings.addons.nutrition.enabledincases.json) for writes to land. A disabled add-on 404s every write (the board surfaces it here as aNot found.tool error) while every GET read stays open, and its 3 nav pages 404 + hide. Enabling is itself a step (§6) — flip it from the board's/addonscatalog (orPATCH /api/addons/nutrition).
This add-on's one process (alongside mcp-bridge-setup's four core servers + two sidecars, and the optional openwhispr / whatsapp add-ons):
| process | what runs | env | bridge port | launchd label | in .mcp.json? |
| nutrition (MCP bridge) | node mcp/nutrition-server/server.mjs via supergateway | CRM_BASE_URL (the board) | 8007 | com.chiefofstaff.mcp-nutrition | yes (http://localhost:8007/mcp) |
Tools (14): food log — log_food, list_food_log, get_food_log, update_food_log,
delete_food_log; pantry — read_pantry, add_pantry_item, update_pantry_item,
remove_pantry_item; meal plan — plan_meal, list_meal_plan, get_meal_plan,
update_meal_plan, remove_meal_plan. The 5 reads (list_*/get_*/read_pantry) are ungated;
the 9 writes are gated. See mcp/nutrition-server/README.md for the full contract.
Machine config comes from the loader (run the preamble in §1): it exports
$REPO_ROOT,$BREW_PREFIX,$NODE_BIN,$SUPERGATEWAY_BIN,$LAUNCH_AGENTS_DIR,$COWORK_CONFIG,$BOARD_URL, and the nutrition keys —$NUTRITION_BRIDGE_PORT(=8007) /$NUTRITION_BRIDGE_URL(=http://localhost:8007) — use those instead of hardcoding paths, the Homebrew prefix, your username, or the port.
Prerequisites
- Node + npm (Homebrew) and the server's deps:
(cd mcp/nutrition-server && npm i). The server imports the sharedpackages/mcp-kitby RELATIVE path, so no workspace install is needed. supergateway(npm install -g supergateway) — the stdio→HTTP bridge for Claude Code (:8007), exactly as in mcp-bridge-setup.- The board reachable at
CRM_BASE_URL(defaulthttp://localhost:3000). The server is a pure wrapper over the board's/api/nutrition/*routes — there is no other backend, no external app, no pairing, no login. (Unlike whatsapp/openwhispr, there is nothing external to install.)
Run the loader preamble as the first line of every shell block below — it exports $REPO_ROOT,
$NODE_BIN, $BREW_PREFIX, $SUPERGATEWAY_BIN, $LAUNCH_AGENTS_DIR, $COWORK_CONFIG,
$BOARD_URL, the bridge port/URL, etc., so nothing below is hardcoded. $U=$(id -u) is derived
inline where launchctl needs it.
source "$(git rev-parse --show-toplevel)/config/load-config.sh"Steps
1. Confirm the config + install the server's deps
The bridge port is a cos.env key (the loader seeds the 8007 default when absent). Confirm it,
and install the server's deps:
source "$(git rev-parse --show-toplevel)/config/load-config.sh"
echo "NUTRITION_BRIDGE_PORT=$NUTRITION_BRIDGE_PORT ($NUTRITION_BRIDGE_URL)" # expect 8007
(cd "$REPO_ROOT/mcp/nutrition-server" && npm i >/dev/null 2>&1)
NUTRITION_BRIDGE_PORTis already inconfig/cos.env.example+config/load-config.sh(default 8007, between whatsapp:8006and the search sidecar:8008). If8007is taken on this machine, change it inconfig/cos.envand re-render §3/§4.
- CHECKPOINT — the port resolves and the server's deps are present:
source "$(git rev-parse --show-toplevel)/config/load-config.sh" test "${NUTRITION_BRIDGE_PORT:-}" = 8007 && echo "port OK (or note your override)" ls "$REPO_ROOT/mcp/nutrition-server/server.mjs" && echo "server present"
2. Verify the stdio server runs
source "$(git rev-parse --show-toplevel)/config/load-config.sh"
CRM_BASE_URL="$BOARD_URL" "$NODE_BIN" "$REPO_ROOT/mcp/nutrition-server/server.mjs" # Ctrl-C
# prints "nutrition MCP server v1 ready (tools: log_food, list_food_log, …; CRM_BASE_URL=…)"- CHECKPOINT — the ready line prints the 14 tool names and echoes the
CRM_BASE_URLit will talk to. (This only proves the server boots; §3 proves the bridge serves it, §7 proves a tool round-trips against the board.)
3. Ensure the .mcp.json nutrition http entry (Claude Code)
The repo already ships the nutrition http entry in $REPO_ROOT/.mcp.json (it was added in an
earlier phase). Confirm it points at the :8007 bridge — do not clobber the other servers:
{ "mcpServers": {
"nutrition": { "type": "http", "url": "http://localhost:8007/mcp" }
}}- CHECKPOINT — the entry is present alongside the core servers:
source "$(git rev-parse --show-toplevel)/config/load-config.sh" grep -q '"nutrition"' "$REPO_ROOT/.mcp.json" && echo ".mcp.json nutrition entry OK".mcp.jsonis GENERATED + CI-checked — never hand-edit it. If the entry is somehow missing, regenerate it from the manifest:node "$REPO_ROOT/scripts/gen-mcp-json.mjs"(seemcp/CLAUDE.md).
4. Install the launchd BRIDGE (:8007)
The plist is generated from mcp/nutrition-server/nutrition.service.json by
scripts/gen-launchd.mjs (see mcp/CLAUDE.md) — it renders the descriptor against the loader's
absolute paths + bridge port (launchd cannot expand $VARS, so the rendered plist carries literal
values) AND does the launchctl bootout → bootstrap → kickstart in one step:
source "$(git rev-parse --show-toplevel)/config/load-config.sh"
node "$REPO_ROOT/scripts/gen-launchd.mjs" --install nutritionThe bridge runs supergateway DIRECTLY around the node child (no launch wrapper — like board/calendar, UNLIKE vault, because this server needs no secret). Its only env is:
PATHstarting with$BREW_PREFIX/bin(launchd can't see an nvm/asdf shim),CRM_BASE_URL=http://localhost:3000(the board; pinned so it doesn't depend on the launchd cwd),COS_DEVICE_ID/COS_DEVICE_ROLE(this machine's identity — the multi-device Devices last-seen signal),COS_MCP_IDLE_EXIT_MS=300000— the idle-exit OPT-IN, set ONLY here (the bridge), never in the direct-stdio Cowork config (see Gotchas).- CHECKPOINT — an MCP
initializeon:8007returnsserverInfo.name == "nutrition":source "$(git rev-parse --show-toplevel)/config/load-config.sh" curl -s -X POST "$NUTRITION_BRIDGE_URL/mcp" \ -H 'Content-Type: application/json' -H 'Accept: application/json, text/event-stream' \ -d '{"jsonrpc":"2.0","id":1,"method":"initialize","params":{"protocolVersion":"2024-11-05","capabilities":{},"clientInfo":{"name":"c","version":"0"}}}' \ | grep -o '"name":"nutrition"' && echo "nutrition bridge OK"
5. Register Cowork (direct stdio)
.mcp.json (§3) covers Claude Code. Claude Cowork Desktop registers differently:
claude_desktop_config.json ($COWORK_CONFIG) accepts stdio command servers only (an HTTP
url is rejected — the same validated rule as mcp-bridge-setup §5). Cowork spawns the node server
itself. Merge the nutrition entry with gen-cowork-config.mjs — it does a backup-first write
that preserves the other servers + preferences and refreshes only the named nutrition entry from
the resolved loader values (secrets inlined; named = additive):
source "$(git rev-parse --show-toplevel)/config/load-config.sh"
node "$REPO_ROOT/scripts/gen-cowork-config.mjs" nutrition- Absolute
nodepath ($NODE_BIN, from the loader): Claude Desktop's spawn env, like launchd's, lacks Homebrew onPATH. - Do NOT set
COS_MCP_IDLE_EXIT_MSin the Cowork entry — Cowork holds one long-lived stdio child for the session; the idle-exit opt-in belongs only in the §4 bridge plist (Gotchas). - Quit + reopen Claude Desktop (⌘Q) — it reads this file only at launch; then
nutritionappears in Cowork's tools. - CHECKPOINT — both registrations present:
source "$(git rev-parse --show-toplevel)/config/load-config.sh" grep -q '"nutrition"' "$REPO_ROOT/.mcp.json" && echo ".mcp.json OK" "$NODE_BIN" -e 'const c=require(process.env.COWORK_CONFIG);process.exit(c.mcpServers&&c.mcpServers.nutrition?0:1)' \ && echo "Cowork config OK"
6. ENABLE the add-on (the gate — writes 404 until you do this)
The plumbing is up, but the add-on ships DISABLED: until you enable it, every WRITE 404s and the
/nutrition nav + pages are hidden. Flip it on — either the /addons UI toggle (with the board
app running) or the API (the toggle the UI calls):
source "$(git rev-parse --show-toplevel)/config/load-config.sh"
curl -s -X PATCH "$BOARD_URL/api/addons/nutrition" \
-H 'Content-Type: application/json' -d '{"enabled":true}' ; echo
# → {"addon":{"id":"nutrition","enabled":true}, "version":<n>}Enabling flips Settings.addons.nutrition.enabled to true in cases.json, bumps db.version →
SSE, so the sidebar's Add-ons nav group + the three /nutrition/* pages light up live (no
reload), and writes start landing.
Enabling nutrition also hard auto-enables the foundational body add-on in the same settings
write (nutrition dependsOn body — the cascade in /api/addons/[id]; body owns the identity,
weight series, and free-text objective that nutrition reads). The /body nav and /api/body
prefix come on with it, but this skill does not wire body's MCP — run /body-mcp-setup
(bridge :8012) next so agents can drive what just turned on:
source "$(git rev-parse --show-toplevel)/config/load-config.sh"
curl -s "$BOARD_URL/api/addons" | node -e 'let s="";process.stdin.on("data",d=>s+=d).on("end",()=>{const a=JSON.parse(s).addons.find(x=>x.id==="body");process.exit(a&&a.enabled===true?0:1)})' && echo "body auto-enabled"- CHECKPOINT — the catalog reports it enabled (and the bridge reachable):
source "$(git rev-parse --show-toplevel)/config/load-config.sh" curl -s "$BOARD_URL/api/addons" | node -e 'let s="";process.stdin.on("data",d=>s+=d).on("end",()=>{const a=JSON.parse(s).addons.find(x=>x.id==="nutrition");process.exit(a&&a.enabled===true?0:1)})' && echo "add-on enabled"
7. End-to-end verify (a tool call round-trips through the board)
Confirm the whole add-on works from the MCP surface, not just the ports — and that the gate is live:
source "$(git rev-parse --show-toplevel)/config/load-config.sh"
# (a) a GATED write round-trips now the add-on is enabled — log_food mints a FOOD-id:
curl -s -X POST "$NUTRITION_BRIDGE_URL/mcp" \
-H 'Content-Type: application/json' -H 'Accept: application/json, text/event-stream' \
-d '{"jsonrpc":"2.0","id":2,"method":"tools/call","params":{"name":"log_food","arguments":{"date":"2026-06-13","slot":"lunch","description":"setup smoke test"}}}' \
| grep -o 'Logged FOOD-[0-9]*' && echo "write round-trips (add-on enabled)"
# (b) an UNGATED read works regardless — list_food_log:
curl -s -X POST "$NUTRITION_BRIDGE_URL/mcp" \
-H 'Content-Type: application/json' -H 'Accept: application/json, text/event-stream' \
-d '{"jsonrpc":"2.0","id":3,"method":"tools/call","params":{"name":"list_food_log","arguments":{"date":"2026-06-13"}}}' \
| grep -o 'Food log' && echo "read works"Delete the smoke-test entry afterwards (delete_food_log with the FOOD-id from (a)), or leave it —
it's harmless. If you call (a) BEFORE §6 you'll see Not found. — that's the gate working, not a
failure; enable the add-on and retry.
- CHECKPOINT —
log_foodreturnsLogged FOOD-<n>,list_food_logreturnsFood log …, and both clients seenutritionafter a ⌘Q reopen. The day-to-day driving (estimating calories, planning a week, linking a meal to the calendar) is then owned by/nutrition-chef— that skill owns the operation; this one only proves the plumbing.
8. (Optional) confirm mcp/ensure-bridges.sh brings it up
nutrition is already wired into mcp/ensure-bridges.sh as an OPTIONAL bridge ("nutrition
${NUTRITION_BRIDGE_PORT:-8007}" in the service list; in the openwhispr|whatsapp|whatsappbridge|nutrition
skip-if-no-plist case), so npm run dev/start brings it up with the rest only when the plist
exists — a machine without the add-on is skipped silently (no WARN … DOWN about :8007). No
edit needed; just confirm:
source "$(git rev-parse --show-toplevel)/config/load-config.sh"
"$REPO_ROOT/mcp/ensure-bridges.sh" | grep -i nutrition # expect: "[mcp] nutrition bridge up on :8007"Manage
launchctl kickstart -k gui/$(id -u)/com.chiefofstaff.mcp-nutrition # restart the bridge
launchctl bootout gui/$(id -u)/com.chiefofstaff.mcp-nutrition # stop the bridge
tail -f mcp/logs/nutrition.err.log # supergateway / server logDisable the add-on (keep the plumbing): flip it off — the nav + pages hide and writes 404 again, but the bridge stays loaded and the data stays on disk + readable:
source "$(git rev-parse --show-toplevel)/config/load-config.sh"
curl -s -X PATCH "$BOARD_URL/api/addons/nutrition" -H 'Content-Type: application/json' -d '{"enabled":false}' ; echo--uninstall (remove the add-on plumbing — the inverse of §3–§6)
Tear the bridge + registrations down (a machine that won't run nutrition). This is the inverse of the install; it does NOT delete data (see the data-purge note below):
source "$(git rev-parse --show-toplevel)/config/load-config.sh"; U=$(id -u)
# 1. Flag the add-on OFF (writes 404, nav hides) — do this first so the surface is dormant:
curl -s -X PATCH "$BOARD_URL/api/addons/nutrition" -H 'Content-Type: application/json' -d '{"enabled":false}' >/dev/null 2>&1 || true
# 2. Stop + remove the launchd bridge:
launchctl bootout gui/$U/com.chiefofstaff.mcp-nutrition 2>/dev/null || true
rm -f "$LAUNCH_AGENTS_DIR/com.chiefofstaff.mcp-nutrition.plist"
# 3. Drop the Cowork entry (backup-first node merge; preserves the other servers + preferences):
"$NODE_BIN" - <<'NODE'
const fs = require('node:fs'), E = process.env, p = E.COWORK_CONFIG;
if (fs.existsSync(p)) {
const cfg = JSON.parse(fs.readFileSync(p,'utf8'));
if (cfg.mcpServers) delete cfg.mcpServers.nutrition;
fs.copyFileSync(p, p + '.bak');
fs.writeFileSync(p, JSON.stringify(cfg, null, 2) + '\n');
console.log('Cowork mcpServers ->', Object.keys(cfg.mcpServers||{}).join(', '));
}
NODE
# 4. Drop the .mcp.json entry (backup-first; keep the other servers):
"$NODE_BIN" - <<'NODE'
const fs = require('node:fs'), E = process.env, p = `${E.REPO_ROOT}/.mcp.json`;
const cfg = JSON.parse(fs.readFileSync(p,'utf8'));
if (cfg.mcpServers) delete cfg.mcpServers.nutrition;
fs.copyFileSync(p, p + '.bak');
fs.writeFileSync(p, JSON.stringify(cfg, null, 2) + '\n');
console.log('.mcp.json mcpServers ->', Object.keys(cfg.mcpServers||{}).join(', '));
NODE
echo "Uninstalled the nutrition bridge + both registrations. ⌘Q + reopen Cowork to drop it from the tool list."The repo SHIPS the
.mcp.jsonnutrition entry and theensure-bridges.shline; removing them from a tracked file leaves a local diff. If you intend to permanently un-ship the add-on (not just this machine), do it in a commit. For a per-machine opt-out, leaving the plist uninstalled is enough —ensure-bridges.shskips nutrition silently when no plist exists, so a stray.mcp.jsonentry pointing at a dead:8007is harmless (Claude Code just shows the server as unreachable).
SEPARATE — purge the data (DESTRUCTIVE, explicit, opt-in). Uninstall leaves the
foodLogs/pantryItems/mealPlanEntriesarrays inboard/data/cases.jsonuntouched (so a re-enable finds your history). To actually erase the nutrition data, edit those three arrays to[]in the store — there is no API to bulk-delete them; do it by hand (or one-by-one viadelete_food_log/remove_pantry_item/remove_meal_plan). The store is backed up daily (the arrays ridecases.jsonin the encrypted off-site backup), so take a fresh backup first if you want a restore point, and know the purge is irreversible once the retention sweep rolls past it. This is intentionally NOT part of--uninstall— disabling/removing the plumbing must never silently destroy a user's food/pantry/meal history.
Gotchas (read before editing)
- The add-on must be ENABLED for writes; reads are always open. The gate is
Settings.addons.nutrition.enabledincases.json(assertAddonEnabledinside the board'smutate()lock). A disabled add-on 404s every WRITE (surfaced as aNot found.tool error) and hides/404s the/nutrition/*pages, but GET reads (list_*,get_*,read_pantry) stay open and the/addonscatalog link stays reachable. So the classic "log_food keeps failing with Not found." is almost always the add-on is off (§6), not a bridge fault — check/api/addons. - The estimation INTELLIGENCE is NOT in the MCP. The tools just store the numbers you give
them; the MCP never estimates calories/macros. That judgment lives in the operator skill
(
/nutrition-chef). Don't expectlog_foodto fill in calories from a description — pass them. COS_MCP_IDLE_EXIT_MSlives ONLY in the bridge plist, never in the Cowork config. mcp-kit's idle-exit is OFF by default so Cowork's long-lived stdio child never dies on idle; the §4 supergateway bridge opts in (300000) to reap supergateway's leaked idle stateless child. Setting it in the Cowork entry would manifest as the dreaded "server transport closed unexpectedly". Same rule as every other bridge (see mcp-bridge-setup → "Why bridges set COS_MCP_IDLE_EXIT_MS").- No sidecar, no external repo, no secret. Unlike whatsapp (Go + Python, external checkout,
QR pairing) and openwhispr (external app store) and vault (Agent SDK +
ANTHROPIC_API_KEY), nutrition is the calendar archetype: an in-repo Nodefetchwrapper that only needs the board onCRM_BASE_URL. There is nothing to clone, build, pair, or authenticate. - The board must be reachable on
CRM_BASE_URL. Every tool is afetchto/api/nutrition/*. If the board app isn't up on:3000(or whereverCRM_BASE_URLpoints), tool calls fail with a connection error — start the board (cd board && npm run dev). - Port
8007must be free (lsof -nP -iTCP:8007 -sTCP:LISTEN). It sits between whatsapp:8006and the search sidecar:8008. A foreign process on it breaks §4 — changeNUTRITION_BRIDGE_PORTincos.envand re-render §3/§4. - launchd cannot expand
$VARS. As with every Cos plist, the rendered plist in~/Library/LaunchAgentscarries literal absolute paths (resolved from the loader byscripts/gen-launchd.mjsin §4) and aPATHthat starts with$BREW_PREFIX/bin— launchd never inherits your login shell and can't see an nvm/asdf shim. The generated plist invokes supergateway by its absolute$BREW_PREFIX/bin/supergatewaypath; the inner--stdio "node …/server.mjs"child resolvesnodevia that$BREW_PREFIX/bin-leadingPATH(supergateway's own#!/usr/bin/env nodeshebang needs it too), which is why thePATHis load-bearing. - The node/simdjson + pm2 gotchas in mcp-bridge-setup apply here too — same Node, same
supergateway, same launchd supervisor. If the bridge dies with a
libsimdjsondyld error after abrew install,brew reinstall node. Don't reintroduce pm2 (pm2 6.x can't fork on this machine — launchd owns the lifecycle).
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.

