agentleFS
Sign inStart free
akasecurityakasecurity/preflight-skills/AGENTS.md

preflight-skills

AGENTS.md1 starsChanged 4 months agoMITView on GitHub

Per-repo conventions for any coding/ops agent. Builds on ~/aka/AGENTS.md (company layer) and the global layer — never repeats them. The public, released multi-model review crew, installed as the preflight plugin. Eight skills, four of them over two zero-dependency scripts: - crew-review — blind cross-family review of a git range. - crew-consult — the same pipeline over a design doc. - biascheck — neutral median authenticity score for a writing draft. - unbias — prompt-only de-slop the session model applies in…

  • 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 repository10 other files it gives its agents.

The file

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

1 rule. About 0 are common, 1 is distinctive, and 0 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.

  • 0

    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. Distinctive:

    No CI or branch protection, so the suite is the only gate.

    line 75

≈ “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 — preflight-skills

Per-repo conventions for any coding/ops agent. Builds on ~/aka/AGENTS.md (company layer) and the global layer — never repeats them.

What this repo is

The public, released multi-model review crew, installed as the preflight plugin. Eight skills, four of them over two zero-dependency scripts:

  • crew-review — blind cross-family review of a git range.
  • crew-consult — the same pipeline over a design doc.
  • biascheck — neutral median authenticity score for a writing draft.
  • unbias — prompt-only de-slop the session model applies in place, outside the engine.
  • secure-research — cited web research fanned out across search engines by scripts/research.mjs.
  • multiplechoice — prompt-only: open decisions and calls made for the user, asked as multiple-choice rounds with a recommendation.
  • experiment — prompt-only: settle a question by a guarded measurement; report the evidence grade.
  • priorart — prompt-only: look for what already exists before building something new.

Public and MIT. Everything committed is visible immediately.

Naming — read this before writing any user-facing string

Renamed from flightcrew-skills on 2026-07-20. The repo is akasecurity/preflight-skills, the plugin id is preflight, the install handle is preflight@akasecurity, and the Homebrew formula is preflight. flightcrew is dead as a public name here.

flightcrew still exists, but it is a different, private, proprietary thing — the engineering harness. Never reintroduce it into this repo's copy.

The old GitHub URL still resolves through a redirect, so a stale link looks healthy. Verify with gh api repos/akasecurity/preflight-skills --jq .full_name rather than by clicking.

GEMINI.md is a product file, not agent instructions

gemini-extension.json sets "contextFileName": "GEMINI.md", so GEMINI.md ships to end users as the Gemini extension's context. It is not the harness-instruction counterpart to this file — read it as product copy, keep it accurate against the shipped skill set, and don't turn it into a pointer. It went stale once already, describing two skills after four had shipped.

CLAUDE.md does point here, as usual.

Structure

  • scripts/crew.mjs — the single engine, importable and executable, only node:* imports. Zero dependencies is a hard constraint, not a preference.
  • scripts/research.mjs — the secure-research engine. It imports crew.mjs's seat runner and is otherwise node:* only as well.
  • workflows/secure-research-workflow.js — the Claude Code adapter for secure-research, shipped through the Claude plugin's default workflows/ scan. A Workflow script can't import modules or spawn processes, so its CLI-engine researchers are Haiku forwarders that run research.mjs. Keep its LENSES, ENGINES and ALL_ENGINES identical to the engine's; tests/workflow.test.mjs checks that.
  • skills/*/SKILL.md — the contracts. These are what the model reads.
  • shared/BIAS.md — the forgiving self-editing catalog unbias applies.
  • shared/TELLS.md — the detection-oriented reference biascheck scores against. Two different catalogs on purpose; don't merge them.
  • tests/*.test.mjs — node --test. tests/stubs/ holds fake claude/codex/agy/grok binaries so the suite never dispatches a real model.

Release surface

A version bump touches more files than it looks: package.json, .claude-plugin/plugin.json, the other four harness manifests (.codex-plugin, .cursor-plugin, .kimi-plugin, gemini-extension.json), CITATION.cff, and CHANGELOG.md. CITATION.cff has been forgotten before. Grep the old version across the tree before tagging.

Downstream copies of the version and name live in akasecurity/marketplace (four files), akasecurity/homebrew-tap (formula + README), and akasecurity/.github (profile/README.md). None of them update on their own.

Workflow

No CI or branch protection, so the suite is the only gate. Run it before pushing to main:

bun run test        # 137 tests; wraps node --test 'tests/*.test.mjs'

The glob matters — node --test tests/ alone picks up the stub binaries in tests/stubs/ and fails.

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