agentleFS
hegeldevhegeldev/hegel-skill/.claude/CLAUDE.md

hegel-skill

CLAUDE.md90 starsChanged 2 months agoMITView on GitHub

This repo is the hegel-skill Agent Skill (distributed via the Agent Skills standard). It teaches coding agents how to write property-based tests with Hegel — a PBT protocol with language-specific libraries (currently Rust, Go, C++, TypeScript, Java, and OCaml), all powered by libhegel, a shared native engine based on Hypothesis and developed in the hegeldev/hegel monorepo alongside the libraries. This repo is content, not code: there is no build, test, or lint step. Edits here directly change agent behavior, so…

The file

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

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

  • 3

    Very specific

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

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

    Every line of skill content is benchmark-driven: it exists because a hegel-skill-bench case showed agents missing bugs without it.

    line 30
  2. Distinctive:

    Language-agnostic: No per-language syntax anywhere.

    line 34
  3. Very specific:

    hegel-review mirrors the bench idiom rubric: When the benchmark's rubric/checklist.json gains a bench-context item, consider mirroring it …

    line 39
  4. Very specific:

    No per-language work is needed: the skill is language-agnostic and points agents at the installed library's own docs and source.

    line 43

≈ “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

This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.

Repository Purpose

This repo is the hegel-skill Agent Skill (distributed via the Agent Skills standard). It teaches coding agents how to write property-based tests with Hegel — a PBT protocol with language-specific libraries (currently Rust, Go, C++, TypeScript, Java, and OCaml), all powered by libhegel, a shared native engine based on Hypothesis and developed in the hegeldev/hegel monorepo alongside the libraries.

This repo is content, not code: there is no build, test, or lint step. Edits here directly change agent behavior, so wording and structure matter.

Layout

.claude-plugin/         # Claude Code marketplace + plugin manifests
skills/hegel/
  SKILL.md              # Entry point — evidence rule, API read-order, technique index
  techniques/
    surfaces.md         # Choosing what to test
    directions.md       # Accept- and reject-direction properties
    generators.md       # Generator breadth: every parameter, hostile inputs
    scale.md            # The scale probe for recursion/complexity bugs
    running.md          # Case counts and run configuration
    triage.md           # Failure ledger, honest reporting
skills/hegel-review/
  SKILL.md              # Checklist for reviewing property-based tests

SKILL.md is a near-empty entry point: API details come from the project's existing hegel tests, the docs, and the installed library source, and each technique file is a short, specific prompt loaded on demand (progressive disclosure). Keep the entry small — new advice goes in a technique file plus one index line, and the index doubles as the pre-stop checklist, which is what makes techniques actually get applied.

Every line of skill content is benchmark-driven: it exists because a hegel-skill-bench case showed agents missing bugs without it. Don't add advice the benchmark hasn't justified.

Conventions When Editing Skill Content

  • Language-agnostic. No per-language syntax anywhere; the skill tells agents to read the library's own docs and source instead.
  • Evidence-based properties. Properties must be grounded in the code under test (names, signatures, docs, existing tests, usage). Don't add advice that encourages inventing properties.
  • Don't weaken generator discipline. Full domains by default is load-bearing — edits that add hedges like "consider narrowing ranges for speed" undo the skill's main lesson. Targeted-subset generators (constructed, not filtered or clamped) are a technique, not a hedge. Resource bounds belong at materialization, never in the drawn domain.
  • One property per test. Preserve this when adding examples.
  • Modify existing test files. The skill tells agents not to create separate files for PBTs. Keep this consistent.
  • hegel-review mirrors the bench idiom rubric. When the benchmark's rubric/checklist.json gains a bench-context item, consider mirroring it here (and vice versa).

Adding a New Language

No per-language work is needed: the skill is language-agnostic and points agents at the installed library's own docs and source. Update the language list in SKILL.md's frontmatter description and in README.md when Hegel ships a new library.

Distribution

  • .claude-plugin/plugin.json — plugin metadata; bump version on user-visible changes.
  • .claude-plugin/marketplace.json — marketplace entry pointing at hegeldev/hegel-skill on GitHub.
  • Installation paths (Claude Code /plugin, Codex $skill-installer, manual) are documented in README.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.