agentleFS
Sign inStart free
srwbswsrwbsw/second-agent-skill/AGENTS.md

second-agent-skill

AGENTS.md0 starsChanged 17 days agoMITView on GitHub

Multi-harness plugin with two entry points, both spawning harness CLIs (gemini, codex, claude, opencode, copilot, qwen, kilo, agy, cmd, cursor, kiro-cli) as subprocesses and embedding diff/file content into the prompt (harnesses don't self-read by default): bin/review.js for read-only second opinions/code review, and its sibling bin/agent.js for delegating an arbitrary engineering task (write tests, fix a bug, refactor) to one of those harnesses inside the repo. Both share internals via bin/lib/. skills/AGENTS.md's native-shortcut block lets the calling assistant skip that subprocess…

  • 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 31,439 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

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

  • 1

    Very specific

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

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

    No build step exists. Never run pnpm build.

    line 28

≈ “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

second-agent (second-agent-skill)

Multi-harness plugin with two entry points, both spawning harness CLIs (gemini, codex, claude, opencode, copilot, qwen, kilo, agy, cmd, cursor, kiro-cli) as subprocesses and embedding diff/file content into the prompt (harnesses don't self-read by default): bin/review.js for read-only second opinions/code review, and its sibling bin/agent.js for delegating an arbitrary engineering task (write tests, fix a bug, refactor) to one of those harnesses inside the repo. Both share internals via bin/lib/. skills/AGENTS.md's native-shortcut block lets the calling assistant skip that subprocess spawn for review.js when the requested harness matches its own host and a native in-session delegation tool is available — see skills/AGENTS.md for details.

Subsystem-specific rules live next to the code:

  • bin/AGENTS.md — runner internals, harness flags, exit codes, secret guard
  • test/AGENTS.md — test conventions + fixture contract
  • skills/AGENTS.md — harness skill docs & the canonical template/checklist source

Agent context files

AGENTS.md is the single source of truth (root + bin/, test/, skills/). Per-tool filenames are symlinks to it so every supported agent picks up the same guidance — edit only AGENTS.md:

  • CLAUDE.md → AGENTS.md (Claude Code; also recognized by Copilot CLI & Cursor)
  • GEMINI.md → AGENTS.md (Gemini CLI; also older Antigravity <1.20.3)
  • QWEN.md → AGENTS.md (Qwen Code)

Codex, opencode, Copilot CLI, Kilo, Cursor, Command Code, and Antigravity ≥1.20.3 read AGENTS.md natively. Gemini/Qwen also get nested AGENTS.md (subdirs) via .gemini/settings.json / .qwen/settings.json (context.fileName: ["AGENTS.md", …]); the root symlinks are the zero-config fallback.

Commands

pnpm run lint          # runs all 8 test suites (safety, shell-quote, env-guard, spawn, answer, agent, locate, host-parity)
pnpm run lint:js       # eslint bin/ test/
pnpm run format        # prettier --write bin/ test/
pnpm run format:check  # prettier --check bin/ test/
  • No build step exists. Never run pnpm build.
  • *.md (README, skills, these files) is NOT prettier-checked — only bin/ and test/.

Versioning & release (do NOT hand-edit versions)

.github/workflows/version-bump.yml runs on push to main and bumps package.json + .claude-plugin/marketplace.json + .codex-plugin/plugin.json from the first line of the head commit (via scripts/bump-version.js), then commits [skip ci]. Mapping: feat:→minor, fix|chore|refactor|perf:→patch, type!:/BREAKING CHANGE→major, else none.

Workflow: small commits → PR to main → squash-merge. The PR title becomes the squash commit's first line, so it must be a conventional commit (it drives the version bump). Husky pre-commit runs prettier+eslint on staged files automatically (no pre-push hook — security reviews are on-demand via the skills).

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