agentleFS
Sign inStart free
mithunycmithunyc/buildloop/AGENTS.md

buildloop

AGENTS.md1 starsChanged 16 days agoMITView on GitHub

This repository builds and maintains buildloop: a small, public skill distribution for evidence-based AI software delivery. Treat this repo as the product and dogfood its own rules. If authorities conflict, stop and report the conflict. Do not smooth it over. Before conclusions or edits: Use the smallest relevant gate set for the touched area: Do not claim success without command output or a clear reason the gate was not applicable. Use templates/ as the source for public contribution artifacts: Root…

  • Where files are headingNew AGENTS.md files have more rules about keeping secrets out of the code, and fewer about code style.

    Of the AGENTS.md files we read, the 31,774 first seen in the last six months, against the older ones: the kind of rule that gained most share, and the kind that lost most. About files like this in general, not this one, and an estimate, since a kind is a pattern in the words.

  • In this repository5 other files it gives its agents.

The file

Show its rules and text

We split this AGENTS.md into the rules it gives an agent

18 rules. About 1 is common, 11 are distinctive, and 6 are very specific to this project.

  • ≈1

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈11

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 6

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    If authorities conflict, stop and report the conflict.

    line 13
  2. Distinctive:

    Before conclusions or edits:

    line 17
  3. Distinctive:

    Run git status --short, git branch --show-current, and git log -1 --oneline.

    line 20
  4. Very specific:

    Read README.md, SECURITY.md, and CONTRIBUTING.md when the task affects public usage.

    line 21
  5. Distinctive:

    Keep edits inside the active task scope and allowed file list.

    line 28
  6. Distinctive:

    Do not start the next task or phase without an explicit instruction.

    line 30
  7. Distinctive:

    If a requested change needs broader scope, stop and justify why.

    line 31
  8. Common:

    Do not publish internal audits, local task state, private scratch docs, secrets, customer data, or machine-local evidence.

    line 36
  9. Distinctive:

    Keep generated artifacts out of the public tree unless they are part of the product contract.

    line 37
  10. Distinctive:

    Use generalized examples. Do not leak private project names or implementation details.

    line 38
  11. Distinctive:

    Use the smallest relevant gate set for the touched area:

    line 42
  12. Very specific:

    Always for manifest/schema changes: node scripts/validate-manifest.mjs

    line 44
  13. Very specific:

    Always for upstream pin changes: node scripts/audit-upstream.mjs

    line 45
  14. Very specific:

    When .buildloop.yml is present: node scripts/gate-runner.mjs

    line 46
  15. Very specific:

    When touching installers: dry-run the relevant installer path, for example powershell -ExecutionPolicy Bypass -File .\scripts\install.ps1 …

    line 47
  16. Distinctive:

    Before completion: git diff --check for touched files and a scoped git diff.

    line 48
  17. Distinctive:

    Do not claim success without command output or a clear reason the gate was not applicable.

    line 50
  18. Very specific:

    Use templates/ as the source for public contribution artifacts:

    line 54

≈ “Common” is an estimate: at least 1 in 5 of the 40,246 AGENTS.md 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
AGENTS.md

AGENTS.md - buildloop Governance

This repository builds and maintains buildloop: a small, public skill distribution for evidence-based AI software delivery. Treat this repo as the product and dogfood its own rules.

Authority Order

  1. Current repo truth: files, git status, latest commit, and command output.
  2. Public product docs: README.md, SECURITY.md, and CONTRIBUTING.md.
  3. Distribution contracts: curated-skills.json, scripts/, schemas/, and tests/.
  4. Templates, examples, playbooks, and reference docs intended for public users.
  5. Prior notes, memory, local state, or external summaries only when verified against this repo.

If authorities conflict, stop and report the conflict. Do not smooth it over.

Bootstrap For Future Agents

Before conclusions or edits:

  1. Read this AGENTS.md.
  2. Run git status --short, git branch --show-current, and git log -1 --oneline.
  3. Read README.md, SECURITY.md, and CONTRIBUTING.md when the task affects public usage.
  4. Inspect the relevant files before editing them.
  5. State assumptions, scope, plan, verification commands, and rollback.

Scope Rules

  • Preserve dirty work. Never revert unrelated changes or other people's edits.
  • Keep edits inside the active task scope and allowed file list.
  • Reuse existing repo patterns before adding new machinery.
  • Do not start the next task or phase without an explicit instruction.
  • If a requested change needs broader scope, stop and justify why.

Public Surface Rules

  • This is a public-facing repo. Keep docs, templates, examples, and scripts safe to publish.
  • Do not publish internal audits, local task state, private scratch docs, secrets, customer data, or machine-local evidence.
  • Keep generated artifacts out of the public tree unless they are part of the product contract.
  • Use generalized examples. Do not leak private project names or implementation details.

Quality Gates

Use the smallest relevant gate set for the touched area:

  • Always for manifest/schema changes: node scripts/validate-manifest.mjs
  • Always for upstream pin changes: node scripts/audit-upstream.mjs
  • When .buildloop.yml is present: node scripts/gate-runner.mjs
  • When touching installers: dry-run the relevant installer path, for example powershell -ExecutionPolicy Bypass -File .\scripts\install.ps1 -Target codex -Mode core -DryRun or bash scripts/install.sh --target codex --mode core --dry-run.
  • Before completion: git diff --check for touched files and a scoped git diff.

Do not claim success without command output or a clear reason the gate was not applicable.

Contribution Standards

Use templates/ as the source for public contribution artifacts:

  • templates/AGENTS.template.md for downstream repo governance.
  • templates/evidence-receipt.md for proof bundles.
  • templates/adversarial-review.md for independent review.
  • templates/buildloop.yml.example for deterministic gates.
  • templates/PRD.md, templates/phase-proposal.md, templates/slice-contract.md, and templates/handoff.md for planning and execution flow.

Root AGENTS.md governs this repo. templates/AGENTS.template.md is a reusable product template for other repos; do not paste it here.

Verification And Reporting

Final reports must include:

  • Exact files changed.
  • Commands run and outcomes.
  • Evidence for key claims.
  • Remaining risks and tradeoffs.
  • Rollback path, usually git restore -- <changed files> before commit.
  • Final verdict: GO, CONDITIONAL_GO, or NO_GO.

Truth over politeness. Evidence over assertion. If you cannot confirm something, say: "I cannot confirm this."

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.

Check your setup, free