agentleFS
CyranoBCyranoB/department-of-prose/CLAUDE.md

department-of-prose

CLAUDE.md2 starsChanged 57 days agoMITView on GitHub

A Claude Code plugin / cross-agent skill family for contextual writing review and revision. Three sibling skills share one versioned editorial catalogue. The optional lexical scorer and rhythm checker produce raw measurements, not authorship verdicts.

The file

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

6 rules. About 0 are common, 1 is distinctive, and 5 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.

  • 5

    Very specific

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

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

    Trigger disambiguation matters: All three skills share vocabulary domain.

    line 30
  2. Very specific:

    Run npm ci && npm run evaluate for deterministic regression coverage.

    line 35
  3. Very specific:

    Read the diff, run bash scripts/link-skills.sh, and try affected skills in a session for changes outside the deterministic scope.

    line 36
  4. Very specific:

    Run python scripts/plugin_version.py bump X.Y.Z with a higher semantic version.

    line 41
  5. Very specific:

    Run python scripts/plugin_version.py check --base origin/main and the packaging checks.

    line 42
  6. Distinctive:

    Commit, PR to main. Both marketplaces pick up the new version on push.

    line 43

≈ “Common” is an estimate: at least 1 in 5 of the 27,641 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

The Department of Prose

A Claude Code plugin / cross-agent skill family for contextual writing review and revision. Three sibling skills share one versioned editorial catalogue. The optional lexical scorer and rhythm checker produce raw measurements, not authorship verdicts.

Structure

  • skills/slop-sense/SKILL.md — the rewriter (36 patterns, workflow, ai;dr mode). Edit this for rewrite-behavior changes.
  • skills/slop-sense/scripts/score.sh — optional algorithmic scorer wrapper around npx slop-detector. Falls back to qualitative-only if Node is missing. Measures lexical tells only (slop words, trigrams, contrast phrases). Reused by slop-check via the sibling path ../slop-sense/scripts/score.sh.
  • skills/slop-sense/scripts/rhythm.py — pure-Python (stdlib only) structural checker for dimensions score.sh does not measure: burstiness, contraction ratio, aphoristic closers, anaphora, raw punctuation counts, and conservative prose cadence candidates. These are editorial heuristics rather than authorship evidence. Reused by slop-check via the sibling path ../slop-sense/scripts/rhythm.py and vendored into the claude.ai bundle.
  • skills/slop-check/SKILL.md — review-only skill (no rewrite). Uses the shared versioned catalogue; allowed-tools deliberately exclude Write/Edit.
  • skills/slop-explain/SKILL.md — per-pattern educational skill. Resolves user request to a pattern number 1-36 via the lookup table, then reads the matching patterns/NN-name.md file.
  • skills/slop-explain/patterns/ — 36 generated per-pattern evidence notes (01-significance-inflation.md through 36-reflexive-formality.md). The versioned catalogue is their policy source.
  • docs/pattern-catalogue-spec.md and docs/pattern-catalogue-source-audit.md — editorial decisions and primary-source limits. scripts/sync-catalogue.py generates the four runtime catalogue copies and 36 evidence notes; --check detects drift.
  • .claude-plugin/plugin.json — plugin manifest. Lists all three skills in skills[] so npx skills@latest can enumerate them together.
  • .claude-plugin/marketplace.json — Claude Code marketplace listing. Mirrors the shared plugin version from plugin.json.
  • .codex-plugin/plugin.json — Codex plugin manifest. Points at the same top-level skills/ directory and includes Codex install-surface metadata.
  • .agents/plugins/marketplace.json — Codex repository marketplace. Exposes the repository root as the department-of-prose plugin.
  • scripts/plugin_version.py — checks or bumps the shared version across both plugin formats and enforces version bumps for plugin changes.
  • scripts/link-skills.sh — dogfooding helper; auto-discovers every SKILL.md under skills/ and symlinks each into ~/.claude/skills (override with SKILLS_DEST).
  • dist-src/slop-sense-bundle/SKILL.md — hand-crafted merged super-skill for the claude.ai / Claude Desktop upload path (one ZIP per skill, so all three modes collapse into one slop-sense skill). Source of truth for the bundle.
  • scripts/package-claude-ai.sh — builds dist/slop-sense.zip from the bundle SKILL.md + vendored score.sh + vendored rhythm.py + vendored patterns/. Asserts exactly 36 pattern files. Output is gitignored.
  • evaluation/ — two-layer quality suite. fixtures/deterministic.json drives executable regressions, coverage.json maps in-scope behaviors, baseline.json tracks strict expected failures, and golden/ holds rubric-reviewed editorial contracts.
  • package.json / package-lock.json — exact slop-detector development dependency used only for reproducible real-scorer integration tests.
  • .github/workflows/evaluation.yml — installs the locked scorer and runs npm run evaluate without model calls.

When making changes

  • Bump the shared plugin version with python scripts/plugin_version.py bump X.Y.Z for any user-visible change. The helper updates both plugin manifests plus the Claude marketplace mirrors; installed clients use this version as the update cache key.
  • The plugin is distributed through Claude Code and Codex marketplaces plus npx skills@latest add CyranoB/department-of-prose (50+ agents). Don't add client-specific paths in any SKILL.md — they have to work cross-agent.
  • Trigger disambiguation matters. All three skills share vocabulary domain. When editing any SKILL.md description, keep the "Prefer X if..." cross-references intact so the agent's router picks the right skill.
  • The scorer is optional by design. Don't make score.sh a hard dependency; the skills must still do useful qualitative analysis when Node is unavailable.
  • The slop-check skill calls the scorer via the sibling path ../slop-sense/scripts/score.sh. If a cross-agent install ever flattens directory structure such that the relative path breaks, vendor a per-skill copy inside slop-check/scripts/ rather than refactoring the path scheme.
  • Pattern numbering (1-36) is the load-bearing convention across all three skills. Change policy in the spec and source audit, then run python3 scripts/sync-catalogue.py; update both lookup tables and packaging rules when IDs change. Re-run bash scripts/package-claude-ai.sh after any catalogue change.
  • Packaging checks use the Python standard library. Run python -m unittest discover -s tests -v, both plugin validators, and bash scripts/package-claude-ai.sh before release.
  • Run npm ci && npm run evaluate for deterministic regression coverage. Expected failures are executable assertions: an unexpected pass fails CI until its exemption is deliberately removed.
  • Read the diff, run bash scripts/link-skills.sh, and try affected skills in a session for changes outside the deterministic scope.

Release flow

  1. Edit SKILL.md files / scripts / deep-dives.
  2. Run python scripts/plugin_version.py bump X.Y.Z with a higher semantic version.
  3. Run python scripts/plugin_version.py check --base origin/main and the packaging checks.
  4. Commit, PR to main. Both marketplaces pick up the new version on push.

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.