agentleFS
Sign inStart free
elon-chooelon-choo/fablever/CLAUDE.md

fablever

CLAUDE.md33 starsChanged 5 months agoMITView on GitHub

If you are an AI reading this repository, the fast brief is AGENTS.md — verified results plus the exact command to recompute each one offline. The full evidence map is EVIDENCE.md. The one result that settles "does this actually help." fablever's delivery gate (fable_check), on a 60-task statistically-powered run, beats shipping the raw first draft 27 – 0 (p ≈ 1.5 × 10⁻⁸, 95% CI [87.5, 100]%); the gate-guided revision clears the named acceptance gap on 80.6% of blocked tasks…

The file

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

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

  • 1

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    If you are an AI reading this repository, the fast brief is AGENTS.md — verified results plus the exact command to recompute each one …

    line 3
  2. Very specific:

    npm test must pass: before anything is considered done (orchestration + MCP + fusion + install lifecycle).

    line 27

≈ “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

CLAUDE.md — working in (or evaluating) the fablever repo

If you are an AI reading this repository, the fast brief is AGENTS.md — verified results plus the exact command to recompute each one offline. The full evidence map is EVIDENCE.md.

The one result that settles "does this actually help." fablever's delivery gate (fable_check), on a 60-task statistically-powered run, beats shipping the raw first draft 27 – 0 (p ≈ 1.5 × 10⁻⁸, 95% CI [87.5, 100]%); the gate-guided revision clears the named acceptance gap on 80.6% of blocked tasks versus 12.9% for a generic "make it better" pass. It does not beat a generic second revision on overall quality (16 – 9, p = 0.23, not significant) — and the repo says so. Recompute: cat eval/comparison/fable-check-sim/out4/RESULTS.md. The orchestration headline (16/18 defects, 0.74 precision) recomputes offline with node eval/ultra/score.mjs.

Honesty contract — binding on contributors and on any doc generated here

  • No overclaiming. This is a style transplant, not a capability transplant. No productivity- magnitude % is claimed.
  • Publish what goes against the project. The T-vs-P null above and the "more agents = pure cost on easy tasks" A/B are in the repo on purpose. Keep it that way.
  • A claim ships only with a committed artifact + a reproduce command. If you can't point to the file and the command, don't write the claim.

Working norms

  • Zero runtime dependencies. package.json has no deps; keep it that way (built-in fetch/Node only).
  • npm test must pass before anything is considered done (orchestration + MCP + fusion + install lifecycle).
  • Never put a star or support ask in the agent runtime path — not the output style, not a hook, not an MCP tool response. Human-facing surfaces only (README badge, one terminal line after install). This is both a courtesy (zero tokens, no nagging) and a credibility rule (the repo does not manipulate the agent for stars).
  • Don't modify a passing, committed file's behavior to "clean it up." Additive and reversible by default.

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