agentleFS
Sign inStart free
Xxx91nXxx91n/anysearch-cli/AGENTS.md

anysearch-cli

AGENTS.md0 starsChanged 4 days agoApache-2.0View on GitHub

AnySearch is an agent CLI (ans) that works inside one domain only. It keeps retrieval results tied to their sources, filters them against a strict domain allowlist, and says there is nothing to answer when no results pass the domain filter. Answers that a provider generates itself (for example Exa or Tavily answer mode) do not pass through this filter and are marked unverified. GLOSSARY.md is the only place to look up what a term means. The rule is kept…

  • Installs packages
  • Commits and pushes
  • 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

2 rules. About 0 are common, 0 are distinctive, and 2 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.

  • 2

    Very specific

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

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

    Before every merge, run the full pnpm test locally and wait for it to pass.

    line 16
  2. Very specific:

    All version control goes through GitButler (but).

    line 27

≈ “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 — anysearch-cli

AnySearch is an agent CLI (ans) that works inside one domain only. It keeps retrieval results tied to their sources, filters them against a strict domain allowlist, and says there is nothing to answer when no results pass the domain filter. Answers that a provider generates itself (for example Exa or Tavily answer mode) do not pass through this filter and are marked unverified. GLOSSARY.md is the only place to look up what a term means.

How to run

  • pnpm install --ignore-scripts — install dependencies. better-sqlite3 ships a prebuilt binary, so no node-gyp.
  • pnpm -r check / pnpm -r test / pnpm -r build — run the same task in every package (turbo).
  • node scripts/ship-gate.mjs — release smoke check. It runs checks, build, packaging, a real install and a real start, and its answer decides whether the release is good.
  • pnpm -C packages/store eval --offline — offline run of the fixed memory test set.
  • cp scripts/hooks/pre-commit-user .git/hooks/pre-commit-user && cp scripts/hooks/pre-push-user .git/hooks/pre-push && chmod +x .git/hooks/pre-commit-user .git/hooks/pre-push — install the test gates. Both run the full workspace test suite. git push and but push run the pre-push hook; a plain git commit runs pre-commit. but commit runs no git hooks at all, so the push gate is the last one that really runs.

Merge gate (#58 direction — decided, see #59)

The rule is kept by discipline, not by a machine. Branch protection is not available (private repo on GitHub Free returns 403), --no-verify can skip the pre-push hook, and a delete may skip the hook entirely. So:

  • Before every merge, run the full pnpm test locally and wait for it to pass. A red CI tells you something, but it does not block anything.
  • pre-push is the last real gate, because it is the only hook that runs on a real push. It is still not a guarantee. A push that passed the hook never replaces a full green run before the merge.
  • We do not make the repo public and do not upgrade the GitHub plan. That option was rejected in #58 and recorded in #59.
  • #67 showed, by pushing a commit that was meant to fail, that the pre-push hook really does block a push unless you pass --no-verify. Proof: the closeout comment on #67.

Directory conventions

  • packages/{kernel,store,retriever,embedding} are shared libraries. apps/{cli,mcp,plugin,dsh-plugin} are the four ways users run the tool. GLOSSARY.md is the word list. docs/agents/ holds notes for agents.

Gotchas (not visible in the code)

  • better-sqlite3@13 loads a prebuilt .node file. allowBuilds: false in pnpm-workspace.yaml is on purpose. Turning it on would need MSBuild + ClangCL. Leave it.
  • pnpm is pinned to 11.24.0 with pmOnFail: error. A version mismatch stops at once and never rewrites the lockfile. Change both pinned fields in one commit.
  • All version control goes through GitButler (but). Never push without an explicit order.
  • Each host's hook format has its own trap. Codex accepts only the hookSpecificOutput shape. Antigravity sends camelCase JSON on stdin, and on PreToolUse an empty {} means DENY. The rest are listed in docs/*-integration.md and in the comments under apps/plugin/src/hooks/adapters/.

Agent skills

Issue tracker

Issues live in GitHub Issues. Use the gh CLI for everything. See docs/agents/issue-tracker.md.

Triage labels

There are five triage roles, and each label string is the role name itself. See docs/agents/triage-labels.md.

Domain docs

This repo has one context: a single GLOSSARY.md at the root. See docs/agents/domain.md.

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