agentleFS
abersonaberson/skill-mesh/skills/lesson-harvest/providers/claude.md

skill-mesh / providers

CLAUDE.md2 starsChanged 2 months agoMITView on GitHub

Claude provider entry point for lesson-harvest; loads the canonical shared core.

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

3 rules. About 0 are common, 2 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    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. Distinctive:

    Use Claude session JSONL, scratchpad identity, Artifact actions, or VS Code deep links only where available and requested by the core.

    line 15
  2. Distinctive:

    Write every record/complete request file into the host's private temporary area or the target repository's own Git metadata area — never …

    line 18
  3. Very specific:

    If a required Claude host tool is unavailable, use the core's documented fallback.

    line 26

≈ “Common” is an estimate: at least 1 in 5 of the 25,680 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

lesson-harvest - Claude entry point

See ../core.md for the full specification.

Provider-specific instructions

  • Load the core in full before acting. Use Claude Code's Skill tool for named-skill calls and Agent/Workflow tools for isolated agents where the core requires those host abstractions.
  • Preserve the core's gates, retry limits, filesystem safety, and exact output contracts.
  • Use Claude session JSONL, scratchpad identity, Artifact actions, or VS Code deep links only where available and requested by the core.
  • Resolve the observation helper from THIS installed package, not from a source checkout and not from the caller's working directory: it is the Python module the core names, in the shared-asset directory that ships one level above this skill's own installed directory. Build an absolute path from this package's location and invoke it with the host's shell tool as python <that absolute path> <command>. If the module is not present beside the installed package, say so in one line and halt with required_tool_missing — never substitute a hand-rolled reader or writer for it, and never edit a saved record by hand.
  • Bind session and source explicitly. session is this Claude Code session's native id when the host exposes one (the session JSONL identity); when it does not, use the literal manual-session — never a guessed, borrowed, or fabricated id. source is the local message or tool-result locator for the evidence in THIS session (for example the assistant turn and the tool result that show the error). Both are inert text: record them, never execute or fetch them.
  • Write every record/complete request file into the host's private temporary area or the target repository's own Git metadata area — never into a tracked worktree path — and delete only the request files this invocation created.

Output normalization

Return only the core's operator-facing report and structured artifacts. Preserve exact locked strings, verdict enums, and exit codes.

Unsupported capabilities

If a required Claude host tool is unavailable, use the core's documented fallback; otherwise halt visibly with required_tool_missing. Never weaken a core gate.

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.