agentleFS
blueprint-conformanceblueprint-conformance/bce/AGENTS.md

bce

AGENTS.md2 starsChanged 6 days agoApache-2.0View on GitHub

This file orients a coding agent (or a human) working on bce itself. If you are instead looking for the snippet to drop into another repo so an agent treats bce as its done-check, that is integrations/AGENTS.md.snippet. bce, the blueprint conformance engine — a fail-closed architecture-conformance gate. You author an EngineeringBlueprint (a durable architectural contract for a repository), and bce measures the code against it: a deterministic conformance score, a fail-closed required-check gate, and hash-chained evidence anyone can re-derive offline.…

  • 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 31,542 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

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

4 rules. About 0 are common, 2 are distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 2

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Describe only machinery that runs: Every workflow, script, and path this repo's docs name must exist or be explicitly marked as a …

    line 28
  2. Distinctive:

    Before you claim a change is done, all required PR checks must be green.

    line 62
  3. Very specific:

    Run node scripts/sync-self-agent-skills.mjs after editing a canonical skill.

    line 78
  4. Very specific:

    Use MCP run_gate {} as the live-tree done-check, or node dist/cli.js gate --repo ..

    line 80

≈ “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 — working in this repository

This file orients a coding agent (or a human) working on bce itself. If you are instead looking for the snippet to drop into another repo so an agent treats bce as its done-check, that is integrations/AGENTS.md.snippet.

What this repository is

bce, the blueprint conformance engine — a fail-closed architecture-conformance gate. You author an EngineeringBlueprint (a durable architectural contract for a repository), and bce measures the code against it: a deterministic conformance score, a fail-closed required-check gate, and hash-chained evidence anyone can re-derive offline. TypeScript/JavaScript framework AST extraction is the mature released path; the current v0.6.2 release retains (from earlier releases) direct TypeScript/JavaScript and structured Python module graphs behind the same language-neutral provider seam, plus declared-stack enforcement over selected package, image, and Node declarations.

The rules that never bend

  1. The gate is the done-check. A change is not done until bce gate is green. A green typecheck or test run is necessary but not sufficient.
  2. On a red, fix the code — never silently edit a blueprint or the baseline. The blueprint is the contract; the baseline only ever shrinks. Changing either is a deliberate, separately-reviewed act.
  3. Fail-closed, no skip flags. There is no --skip / --no-verify / --force. If the engine cannot honestly grade, it refuses (exit 2) — treat a refusal as a red, never a pass. Adoption is handled by committed config (advisory mode, shrink-only baseline), never an invisible flag.
  4. Nothing internal leaks. A dependency-free leakage-gate runs from commit #1 and rejects any internal identifier, host, or credential-shaped string. Keep it green.
  5. Describe only machinery that runs. Every workflow, script, and path this repo's docs name must exist or be explicitly marked as a forward/deferred item. No fictional pipelines.

Map of the repository

PathWhat it is
src/the engine (extraction → evaluate → score → verdict); src/extractors.ts is the only ts-morph importer
src/cli.tsthe bce CLI: lifecycle, proposal/review, extraction, gate, evidence, and portfolio commands; bce --help is authoritative
src/mcp-server.tsthe bce-mcp THIN stdio server (logic-free shell over the engine API)
spec/SPEC.mdthe normative specification (artifact model, taxonomy, scoring, exit codes §13, report contract §11)
spec/schemas/published JSON Schemas — generated, never hand-edited
spec/conformance-vectors/input→expected-verdict vectors AS DATA (no runner, no levels at v1alpha1)
docs/quickstart, adopt-existing-repo, agent-loop, exit-codes, report-contract, self-hosting, evidence-format, faq
examples/quickstart/the guaranteed offline RED→fix→GREEN walkthrough
integrations/drop-in agent snippets (CLAUDE.md / AGENTS.md / .cursorrules) + MCP details
skills/the Agent Skill — the author → validate → run → teeth → gate lifecycle as a folder-per-skill SKILL.md, for an agent creating a contract rather than working inside one
prompts/the experimental blueprint-author pack + its 5-public-repo validation
fixtures/ + corpus/the seeded-defect corpus (recall denominator) and its machine-readable manifest
.blueprints/this repo's OWN blueprint — bce gates itself (see docs/self-hosting.md)
.github/workflows/ci, leakage-gate, self-gate, docs-site-check, and active publish-schemas (all fail-closed)
.github/labels.json + docs/maintainer-operations.mdcanonical one-human contributor queue; read-only response audit, no phantom approver
scripts/build-docs-site.mjsassembles _site/ — the schemas at their $id paths + the rendered docs site; dependency-free, fail-closed on IA drift and dangling links

Build, test, gate (Node 22 toolchain)

npm ci            # clean, lockfile-exact install
npm run build     # tsup + declaration emit → dist/
npm run typecheck # tsc --noEmit, full tree
npm test          # the full vitest suite
node dist/cli.js gate --repo . --repo-name blueprint-conformance/bce   # gate this repo

Before you claim a change is done, all required PR checks must be green. For release or public-site work, also wait for all release-relevant post-merge workflows and all six portability legs on the exact merged main commit; a passing PR or Pages deployment is not completed release verification. See the portability contract and CONTRIBUTING.md for the DCO sign-off and agent-authored-PR conventions (gate green + evidence ref in the PR body), and docs/self-hosting.md for the self-gate and the admin-override incident policy.

Machine-readable index

llms.txt is the compact, link-first index of this repository for an agent that prefers a flat map. It points at the same documents as the table above.

Self-adoption integration

Project skills are installed under .agents/skills/ from the canonical skills/ tree. Run node scripts/sync-self-agent-skills.mjs after editing a canonical skill; CI checks freshness. Build first, then the project MCP configuration runs node dist/mcp-server.js from this checkout. Use MCP run_gate {} as the live-tree done-check, or node dist/cli.js gate --repo .. Review .bce-governance.json for the explicit solo-steward authority. A self-ratified decision claims no independent review. Changes to policy, agent configuration, or installed skills require the human owner's review.

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.