agentleFS
bazaarjapanbazaarjapan/codex-autoloop-template/.agents/skills/autoloop/SKILL.md

autoloop

Skill0 starsChanged 3 months agoMITView on GitHub

Run this repository's bounded specification-driven development loop from product intake through specification, implementation, GitHub Issue, pull request, CI repair, independent Codex review, and safe auto-merge. Use when the user invokes $autoloop or asks to build a product, feature, or substantial behavior autonomously in this template.

1 other project uses a version of this.

Adopted by

Most recently taken up by

  • Other versions1 project uses a version of this.

    Word for word · 0

    None found.

    Edited · 0

    None found.

    Wrote their own version · 1

    1. sameerhimati/claude-code-kit★ 1same name

    How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.

  • In this repository3 other files it gives its agents.

The file

Show its rules and text

We split this skill into the rules it gives an agent

17 rules. About 0 are common, 9 are distinctive, and 8 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈9

    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
  • could be automatic: a setting could enforce it
  1. Very specific:

    Read AGENTS.md, .autoloop/config.json, .autoloop/state.json, and .autoloop/policies/*.md completely.

    line 12
  2. Very specific:

    Run python scripts/autoloop/autoloop.py doctor and python scripts/autoloop/autoloop.py validate.

    line 13
  3. Distinctive:

    If the state has an active unfinished feature, resume it.

    line 15
  4. Distinctive:

    Read only the phase reference needed next:

    line 16
  5. Very specific:

    Use scripts/autoloop/autoloop.py for all state transitions, counters, evidence, and specification locking.

    line 24
  6. Distinctive:

    Do not merge in the blocked state.

    line 39
  7. Very specific:

    Specification: - create the five artifacts under specs/<slug>/, critique them independently, repair findings, validate, and lock the digest.

    line 44
  8. Very specific:

    Specification delivery: - when GitHub is configured, put the locked specification on a dedicated spec/<slug> branch and managed pull …

    line 45
  9. Distinctive:

    Issue: - create one implementation Issue from the locked specification and store its URL or number as evidence.

    line 46
  10. Very specific:

    Implementation: - create feat/<issue>-<slug> and implement tasks against the acceptance criteria and locked traceability plan.

    line 47
  11. Very specific:

    Local verification: - run validate, verify, and repository tests.

    line 48
  12. Very specific:

    Do not begin product implementation before SPEC_LOCKED.

    line 54
  13. Distinctive:

    Do not rewrite locked specification files from an implementation branch.

    line 55
  14. Distinctive:

    Do not modify protected control paths in a managed feature pull request.

    line 56
  15. Distinctive:

    Do not weaken, skip, delete, or suppress a gate to make it pass.

    line 57
  16. Distinctive:

    Do not expose secrets.

    line 60
  17. Distinctive:

    Do not claim completion without evidence tied to the latest or merged commit.

    line 61

≈ “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.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Codex AutoLoop

Turn a product goal into a locked specification and verified implementation. Continue autonomously after intake while evidence passes; stop closed as AUTOMATION_BLOCKED if a limit or safety boundary is reached.

Start

  1. Read AGENTS.md, .autoloop/config.json, .autoloop/state.json, and .autoloop/policies/*.md completely.
  2. Run python scripts/autoloop/autoloop.py doctor and python scripts/autoloop/autoloop.py validate. Use python3 if required by the host.
  3. Inspect the repository and existing product before proposing behavior. Preserve unrelated user changes.
  4. If the state has an active unfinished feature, resume it. Do not start a second feature over it.
  5. Read only the phase reference needed next:
    • intake and scope: references/intake.md
    • specification and critique: references/specification.md
    • implementation and local convergence: references/implementation.md
    • Issue, pull requests, review loops, and merge: references/shipping.md

State discipline

Use scripts/autoloop/autoloop.py for all state transitions, counters, evidence, and specification locking. Do not hand-edit .autoloop/state.json to bypass validation.

After every phase:

  1. validate the repository;
  2. record current evidence;
  3. transition only through an allowed edge;
  4. continue immediately if the next step is authorized and safe.

On failure, diagnose and repair within the corresponding configured loop limit and total runtime. Increment the counter before retrying. If a limit is exhausted or the same cause repeats beyond its limit, run:

python scripts/autoloop/autoloop.py block --reason "concise evidence-backed reason"

Do not merge in the blocked state.

Required lifecycle

  1. Intake - obtain or infer enough information to specify observable behavior. Questions gather requirements, not approval.
  2. Specification - create the five artifacts under specs/<slug>/, critique them independently, repair findings, validate, and lock the digest.
  3. Specification delivery - when GitHub is configured, put the locked specification on a dedicated spec/<slug> branch and managed pull request; let required checks and auto-merge complete.
  4. Issue - create one implementation Issue from the locked specification and store its URL or number as evidence.
  5. Implementation - create feat/<issue>-<slug> and implement tasks against the acceptance criteria and locked traceability plan.
  6. Local verification - run validate, verify, and repository tests. Repair within bounds.
  7. Implementation delivery - open a managed pull request that closes the Issue, wait for current-head CI and the native Codex review gate, repair actionable findings, and re-run all evidence.
  8. Merge - arm GitHub auto-merge only through the included workflow. Confirm the merged commit, then transition through MERGED to DONE and close the Issue if it was not closed automatically.

Non-negotiable rules

  • Do not begin product implementation before SPEC_LOCKED.
  • Do not rewrite locked specification files from an implementation branch.
  • Do not modify protected control paths in a managed feature pull request.
  • Do not weaken, skip, delete, or suppress a gate to make it pass.
  • Require codex-review-gate evidence from a native Codex review of the latest head; do not accept a stale review from an older commit.
  • Do not force-push.
  • Do not expose secrets.
  • Do not claim completion without evidence tied to the latest or merged commit.

Handoff

For DONE, report the goal, locked-spec path, Issue, pull requests, merged commit, tests and gates, and any recorded assumptions. For AUTOMATION_BLOCKED, report the exact phase, failure evidence, attempts used versus limit, and the smallest safe action needed to resume.

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.