bootstrap-dogfood-project
Skill9 starsChanged 7 months agoMITView on GitHub
Wire a downstream/dogfood app repo into the hub's feedback loop and VERIFY all five preconditions are green before the app is declared "set up". Closes the silent loophole where an app improves locally but nothing flows back to the hub. Use right after creating/synthesizing a new app repo that should feed patterns + telemetry back to this hub, before its first feature commit.
In this repository44 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- anthropic-multi-agent-research-system-skillSkill
- apply-selectionsSkill
- brainstormSkill
- branch-choiceSkill
- cc-adoption-scoutSkill
- claude-guardianSkill
- continueSkill
- end-sessionSkill
- executing-plansSkill
- five-advisorsSkill
- get-work-doneSkill
- git-branch-lifecycleSkill
- githubSkill
- grill-meSkill
- hook-transcript-fixture-testSkill
- instagramSkill
- linkedinSkill
- modern-data-pipeline-diagramSkill
- pattern-qualitySkill
- platform-event-live-probeSkill
- plugin-lifecycleSkill
- promote-to-coreSkill
- prompt-auto-enhanceSkill
- provision-reportSkill
- redditSkill
- review-new-claude-featuresSkill
- save-sessionSkill
- scan-discovery-reportSkill
- scan-repoSkill
- scan-urlSkill
- self-improveSkill
- skill-evaluatorSkill
- skill-masterSkill
- ssot-workflow-auditSkill
- start-sessionSkill
- synthesize-hubSkill
- synthesize-projectSkill
- twitter-xSkill
- verify-effect-at-destinationSkill
- workflow-doc-reviewerSkill
- writing-plansSkill
- writing-skillsSkill
- youtube-transcriptSkill
The file
We split this skill into the rules it gives an agent
9 rules. About 0 are common, 1 is distinctive, and 8 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.
8
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 47
Verify the app has a pushed GitHub remote: gh repo view <owner/name>.
- Very specific: line 54
Read config/repos.yml; if the app slug is absent, add its entry (match the existing schema of neighboring entries — do not invent fields).
- Very specific: line 72
Ensure the app's .claude/synthesis-config.yml sets allow_hub_sharing: true (required for /synthesize-hub to pull its synthesized: true …
- Very specific: line 105
MUST verify all five preconditions.
- Very specific: line 106
MUST treat a gitignored or untracked .claude/learnings.json as a hard fail — it is the silent-loophole root cause.
- Distinctive: line 107
MUST NOT silently create/push the app's GitHub repo — that outward action is escalated to the user with a one-line instruction.
- Very specific: line 108
MUST match the existing config/repos.yml entry schema when enrolling — do not invent fields.
- Very specific: line 109
MUST run only from the hub repo (where config/repos.yml lives).
- Very specific: line 110
MUST re-run cleanly (idempotent): an already-wired repo verifies READY without duplicate edits.
≈ “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
/bootstrap-dogfood-project — wire + verify the project→hub feedback loop
This is a hub-only operational skill: it edits the hub's config/repos.yml
and inspects a downstream app repo. It is NOT distributed to projects (it has no
meaning inside a consumer project). Its STEP 6 gate is the deterministic
replacement for the prose checklist in plans/idea-to-deploy-readiness.md —
the app is NOT "set up" until all five preconditions verify green.
Why it exists: the four project→hub feedback flows (/contribute-practice,
collate.py, aggregate_telemetry.py, /synthesize-hub) each assume
preconditions that, if unset at project birth, silently sever the loop — the app
gets better locally and the hub learns nothing. This skill sets them and refuses
to pass until they hold.
Input: $ARGUMENTS — the app repo (owner/name for a GitHub repo, or a
local path). --verify-only checks without mutating.
STEP 1: INIT
- Resolve the hub root (
git rev-parse --show-toplevelfrom this repo) and confirmconfig/repos.ymlexists — if not, this is not the hub; STOP. - Resolve the app repo: GitHub slug
owner/name, or a local path. Capture both the slug and (if local) the working dir. - Initialize a verdict accumulator: five checks, each
unset → pass|fail.
STEP 2: GitHub remote (the app must be API-visible)
The hub's aggregators (collate.py, aggregate_telemetry.py) read the app via
the GitHub API — a local-only repo is invisible.
- Verify the app has a pushed GitHub remote:
gh repo view <owner/name>. - If missing: creating/pushing a repo is an outward action — BLOCK this
check with the one-line instruction to
gh repo create+ push, and recordgithub_remote: fail. Do not fabricate the repo silently.
STEP 3: Enroll in the hub's config/repos.yml
- Read
config/repos.yml; if the app slug is absent, add its entry (match the existing schema of neighboring entries — do not invent fields). --verify-only→ only report presence, do not edit.- Record
enrolled: pass|fail.
STEP 4: Committed, non-ignored .claude/learnings.json
The hub reads the committed learnings file; a gitignored one emits ZERO signal (loop-engineering's #1 silent failure).
- In the app repo: confirm
.claude/learnings.jsonexists (seed{"learnings": []}if not), is NOT matched by any.gitignorerule, and is tracked by git (git ls-files --error-unmatch .claude/learnings.json). - A gitignored or untracked learnings file is a hard fail — record
learnings_committed: failwith the exact reason.
STEP 5: allow_hub_sharing: true in the app's .claude/synthesis-config.yml
- Ensure the app's
.claude/synthesis-config.ymlsetsallow_hub_sharing: true(required for/synthesize-hubto pull itssynthesized: truepatterns). - Record
hub_sharing: pass|fail.
STEP 6: Routing discipline declared
- Confirm the app's
CLAUDE.md(or a rule) states thelearnings-routing.mdsplit: GENERIC learnings flow hub-ward, PRODUCT-SPECIFIC stay local. Add a one-line pointer if absent. - Record
routing_declared: pass|fail.
STEP 7: VERIFY GATE (the deterministic close)
Re-check all five. Emit a verdict and refuse to declare ready on any fail:
============================================================
Dogfood bootstrap: <READY | BLOCKED>
app: <owner/name>
[✓|✗] github_remote [✓|✗] enrolled (config/repos.yml)
[✓|✗] learnings_committed [✓|✗] hub_sharing
[✓|✗] routing_declared
Blockers: <list or none>
============================================================READY only when all five are ✓. On any ✗, output BLOCKED with the exact
fix per blocker and STOP — the feedback loop is not wired until they are green.
CRITICAL RULES
- MUST verify all five preconditions; MUST output
BLOCKED(neverREADY) if any fails. - MUST treat a gitignored or untracked
.claude/learnings.jsonas a hard fail — it is the silent-loophole root cause. - MUST NOT silently create/push the app's GitHub repo — that outward action is escalated to the user with a one-line instruction.
- MUST match the existing
config/repos.ymlentry schema when enrolling — do not invent fields. - MUST run only from the hub repo (where
config/repos.ymllives); this skill is hub-only and MUST NOT be distributed to consumer projects. - MUST re-run cleanly (idempotent): an already-wired repo verifies
READYwithout duplicate edits.
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.

