agentleFS
Sign inStart free
WenjixWenjix/quorum/AGENTS.md

quorum

AGENTS.md2 starsChanged 5 months agoMITView on GitHub

Operational guidance for humans and coding agents working in this repo. - Registry: npmjs, pinned in .npmrc. Do not switch to registry.npmmirror.com or commit a lockfile that resolves from it — its advisory endpoint returns NOT_IMPLEMENTED, so npm audit silently can't run. - The zod override in package.json is an intentional security pin. zod is a peer dependency @cursor/sdk@1.2.0 declares as ^3.25.0 || ^4.0.0; without the pin npm installs the 3.x line, which carries an open advisory fixed only in…

  • Reads credentials
  • 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 repositoryOne other file it gives its agents.

The file

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

3 rules. About 0 are common, 0 are distinctive, and 3 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.

  • 3

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    After editing src/, rebuild and invoke with node dist/src/cli.js <command> (no npm link step).

    line 31
  2. Very specific:

    Run artifacts are written under .quorum/ (gitignored).

    line 57
  3. Very specific:

    Use --no-canvas-mirror on offline/CI runs so the runner doesn't write into the Cursor managed canvases directory under ~/.cursor.

    line 62

≈ “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

Agent & contributor notes

Operational guidance for humans and coding agents working in this repo.

Dependencies / npm

  • Registry: npmjs, pinned in .npmrc. Do not switch to registry.npmmirror.com or commit a lockfile that resolves from it — its advisory endpoint returns NOT_IMPLEMENTED, so npm audit silently can't run.
  • The zod override in package.json is an intentional security pin. zod is a peer dependency @cursor/sdk@1.2.0 declares as ^3.25.0 || ^4.0.0; without the pin npm installs the 3.x line, which carries an open advisory fixed only in 4.6.6. Do not remove or loosen it, and do not regenerate the lockfile against another registry. The former undici, tar and @tootallnate/once pins went away when the SDK dropped connect-node and sqlite3. Rationale, verification, and the removal condition live in docs/dependencies.md (tracked in #5).
  • Auditing: npm audit --audit-level=moderate (npmjs registry required).

Build, test & run

Single Node.js/TypeScript CLI — no server, no watch mode.

  • Commands: npm run build (tsc), npm test (builds, then node --test dist/test/*.test.js, then the python suite under scripts/tests with python3 -m unittest discover -s scripts/tests -v), npm run typecheck. build/test clean dist/ first (prebuild), so stale compiled output from a branch switch can't leak into a run.
  • Skill scripts (scripts/*.py): they must import on Python 3.9, so put from __future__ import annotations in any module that uses PEP 604 unions (X | None). scripts/quorum_ledger.py is the only writer of the PR comment: every other path (the CLI's synthesize, the skill, an agent following the embedded protocol) posts through it.
  • After editing src/, rebuild and invoke with node dist/src/cli.js <command> (no npm link step).
  • Offline / no-credentials runs: plan-pr <PR-URL> is plan-only by design — it builds the DAG → state → report → Canvas without calling any cloud backend. It still reads the PR's findings from GitHub unless you pass them locally with --scored <clusters.scored.json>, which makes the run fully offline: node dist/src/cli.js plan-pr https://github.com/OWNER/REPO/pull/N --scored clusters.scored.json. On the lower-level explore/triage-pr commands the same cloud-skip is the --plan-only flag.
  • Live commands — run-pr, post-pr, run-dag, and triage-pr without --plan-only — call the exploration backend and/or GitHub. They need the selected backend's key — CURSOR_API_KEY (default) or ANTHROPIC_API_KEY (with --provider anthropic / QUORUM_PROVIDER=anthropic) — plus an authenticated gh CLI. The Anthropic backend also fetches the PR diff for code context, so GITHUB_TOKEN/gh auth improves it but is non-fatal if missing. Without credentials the runner still completes, recording per-task ERROR/SKIPPED instead of crashing.
  • Persisting API keys — quorum auth saves CURSOR_API_KEY / ANTHROPIC_API_KEY / QUORUM_PROVIDER to ~/.config/quorum/credentials.json (0o600) so you don't export them every shell. Key resolution order: --api-key flag > process env > config file. Override the file path with QUORUM_CONFIG.
  • Repackaging the skill: quorum.skill is the committed, installable zip of SKILL.md, scripts/ and references/. The README's install steps unzip it, so a stale zip ships stale scripts. After changing any of those files: commit the change, run npm run build:skill (it packages from HEAD via git archive, so uncommitted edits are left out), and commit the refreshed zip. .gitattributes pins the packaged files to LF so Windows core.autocrlf can't inject CRLF (exec bits come from the index), and marks scripts/tests/** and __pycache__ export-ignore so the zip carries no tests. The npm tarball leaves tests out too: the negated !scripts/tests/ and !scripts/**/__pycache__/ entries in package.json files do the work (a root .npmignore alone does not override the files whitelist), and .npmignore repeats them. Check with npm pack --dry-run.
  • Run artifacts are written under .quorum/ (gitignored).

Cursor Cloud

  • The cloud-agent environment runs npm install on startup (configured in Cursor, not in the repo).
  • Use --no-canvas-mirror on offline/CI runs so the runner doesn't write into the Cursor managed canvases directory under ~/.cursor.

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