agentleFS
astriaaiastriaai/skills/CLAUDE.md

skills

CLAUDE.md2 starsChanged 26 days agoMITView on GitHub

The canonical public skills live in top-level <name>/SKILL.md directories. For database-backed workflows, the full recipe is owned by the private sdbooth/doc/skills/ source and published through agent_skills:sync. Public mux and fashion-music-track skills are loaders only: preserve their names and triggers, and read their instructions with get_skill. Change recipes in the private source; do not copy them into the public skills or packaging. Source manifests and assets live in packaging/; generated packages and marketplaces live under ignored dist/marketplace/ and ship as release…

The file

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

3 rules. About 0 are common, 1 is distinctive, and 2 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.

  • 2

    Very specific

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

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

    Never commit generated plugin copies.

    line 41
  2. Distinctive:

    No SessionStart CLI-auth hook: OAuth is managed by the host.

    line 43
  3. Very specific:

    Keep skill frontmatter descriptions focused.

    line 44

≈ “Common” is an estimate: at least 1 in 5 of the 26,980 CLAUDE.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
CLAUDE.md

Astria skills — repository guide

The canonical public skills live in top-level <name>/SKILL.md directories. For database-backed workflows, the full recipe is owned by the private sdbooth/doc/skills/ source and published through agent_skills:sync. Public mux and fashion-music-track skills are loaders only: preserve their names and triggers, and read their instructions with get_skill. Change recipes in the private source; do not copy them into the public skills or packaging. Source manifests and assets live in packaging/; generated packages and marketplaces live under ignored dist/marketplace/ and ship as release ZIPs. The Astria web-agent sandbox also bundles these skills with its independently authenticated CLI and private embedded skills.

API transport

Installed plugins use OAuth-connected Astria MCP tools for supported API operations. The host owns login, token storage and refresh. Use live tool schemas; names may have a host namespace. Never read host credentials or silently fall back to the CLI after an MCP authentication failure.

The CLI remains available for local files, downloads, video inspection/Variate, workspace/template creation, landing pages, board edits and agent handoff. It has separate authentication. The embedded Astria web-agent sandbox may expose only the authenticated CLI; preserve that supported environment. Conditional instructions live in astria-api/references/cli.md, not in the default MCP entrypoint. Never build a second API client in skills.

bin/astria is vendored from astriaai/cli. Change the canonical CLI there, then run scripts/sync-cli.sh [ref]. Shared operation definitions and argparse generate MCP schemas; add new remote-safe parameters there and synchronize Rails through bin/sync-astria-mcp.

Packaging

  • packaging/marketplaces/claude.json owns the canonical skill list. Add each skill as a "./<name>" path.
  • mcp.json owns the portable HTTPS endpoint. scripts/sync-plugin.py generates root and packaged .mcp.json HTTP compatibility configs.
  • scripts/sync-plugin.sh --no-local-links copies canonical skills (including references), manifests, MCP configuration and the vendored CLI to dist/marketplace/plugins/. Without that flag it also refreshes developer symlinks.
  • Never commit generated plugin copies. Build --target marketplace for the complete installable marketplace ZIP; individual targets remain upload ZIPs.
  • No SessionStart CLI-auth hook: OAuth is managed by the host.
  • Keep skill frontmatter descriptions focused; retain shell permissions only for skills that actually need optional terminal workflows.
  • Public skills must not depend on private web-chat browser command protocols. Composer tools apply only when the host exposes them.

Validation

scripts/sync-plugin.sh --no-local-links
python3 scripts/validate-plugin.py
python3 scripts/build-plugin.py /tmp/astria-plugin-preview

The validator checks source/package parity, MCP wiring and skill JSON tool examples against the vendored CLI's generated schemas. Also run relevant Rails MCP specs if changing shared operation behavior. Deploying the endpoint, registering OAuth clients and publishing a directory version are separate release steps; see docs/PUBLISHING.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.