Skip to content

netresearch/agent-rules-skill / skills/agent-rules

agent-rules

Flagged

Use when creating or updating AGENTS.md files, .github/copilot-instructions.md, or other AI agent rule files, onboarding AI agents to a project, standardizing agent documentation, or when anyone mentions AGENTS.md, agent rules, project onboarding, or codebase documentation for AI agents.

FileKindSize
AGENTS.mdmarkdown4.0 KB
CLAUDE.mdmarkdown9 B
SKILL.mdskill md5.2 KB
assets/.markdownlint-cli2.jsoncother624 B
assets/example-workflows/validate-agents.ymlconfig2.8 KB
assets/root-thin.mdmarkdown5.7 KB
assets/root-verbose.mdmarkdown5.5 KB
assets/scoped/backend-go.mdmarkdown3.6 KB
assets/scoped/backend-php.mdmarkdown3.0 KB
assets/scoped/backend-python.mdmarkdown2.6 KB
assets/scoped/backend-typescript.mdmarkdown2.8 KB
assets/scoped/claude-code-skill.mdmarkdown3.4 KB
assets/scoped/cli.mdmarkdown2.7 KB
assets/scoped/concourse.mdmarkdown5.2 KB
assets/scoped/ddev.mdmarkdown2.4 KB
assets/scoped/docker.mdmarkdown5.3 KB
assets/scoped/documentation.mdmarkdown3.5 KB
assets/scoped/examples.mdmarkdown3.4 KB
assets/scoped/frontend-typescript.mdmarkdown2.7 KB
assets/scoped/github-actions.mdmarkdown5.2 KB
assets/scoped/gitlab-ci.mdmarkdown4.6 KB
assets/scoped/oro-bundle.mdmarkdown6.5 KB
assets/scoped/oro-project.mdmarkdown5.5 KB
assets/scoped/python-modern.mdmarkdown6.1 KB
assets/scoped/resources.mdmarkdown3.5 KB
assets/scoped/skill-repo.mdmarkdown5.4 KB
assets/scoped/symfony.mdmarkdown5.3 KB
assets/scoped/testing.mdmarkdown3.0 KB
assets/scoped/typo3-docs.mdmarkdown3.6 KB
assets/scoped/typo3-extension.mdmarkdown5.1 KB
assets/scoped/typo3-project.mdmarkdown4.7 KB
assets/scoped/typo3-testing.mdmarkdown3.0 KB
checkpoints.yamlconfig14.4 KB
evals/evals.jsonconfig22.9 KB
references/ai-contribution-guidelines.mdmarkdown2.6 KB
references/ai-tool-compatibility.mdmarkdown8.0 KB
references/directory-coverage.mdmarkdown2.5 KB
references/examples/coding-agent-cli/AGENTS.mdmarkdown2.5 KB
references/examples/coding-agent-cli/go.modmanifest45 B
references/examples/coding-agent-cli/scripts-AGENTS.mdmarkdown10.1 KB
references/examples/express-api-ts/.env.exampleother232 B
references/examples/express-api-ts/AGENTS.mdmarkdown3.1 KB
references/examples/express-api-ts/package.jsonmanifest757 B
references/examples/express-api-ts/pnpm-lock.yamlmanifest101 B
references/examples/express-api-ts/src/AGENTS.mdmarkdown3.1 KB
references/examples/express-api-ts/src/config.tsscript917 B
references/examples/express-api-ts/src/controllers/userController.tsscript2.1 KB
references/examples/express-api-ts/src/index.tsscript671 B
references/examples/express-api-ts/src/middleware/errorHandler.tsscript1.0 KB
references/examples/express-api-ts/src/middleware/requestLogger.tsscript471 B
references/examples/express-api-ts/src/routes/health.tsscript428 B
references/examples/express-api-ts/src/routes/users.tsscript397 B
references/examples/express-api-ts/src/utils/errors.tsscript1.2 KB
references/examples/express-api-ts/src/utils/logger.tsscript441 B
references/examples/express-api-ts/tsconfig.jsonconfig630 B
references/examples/fastapi-app/.env.exampleother348 B
references/examples/fastapi-app/AGENTS.mdmarkdown3.2 KB
references/examples/fastapi-app/pyproject.tomlmanifest1.8 KB
references/examples/fastapi-app/src/AGENTS.mdmarkdown2.9 KB
references/examples/fastapi-app/src/__init__.pyscript58 B
references/examples/fastapi-app/src/config.pyscript1.1 KB
references/examples/fastapi-app/src/main.pyscript1.7 KB
references/examples/fastapi-app/src/models/__init__.pyscript278 B
references/examples/fastapi-app/src/models/item.pyscript1.0 KB
references/examples/fastapi-app/src/models/user.pyscript922 B
references/examples/fastapi-app/src/routes/__init__.pyscript110 B
references/examples/fastapi-app/src/routes/health.pyscript470 B
references/examples/fastapi-app/src/routes/items.pyscript1.9 KB
references/examples/fastapi-app/src/routes/users.pyscript1.7 KB
references/examples/fastapi-app/src/services/__init__.pyscript174 B
references/examples/fastapi-app/src/services/item_service.pyscript2.5 KB
references/examples/fastapi-app/src/services/user_service.pyscript2.1 KB
references/examples/fastapi-app/uv.lockmanifest136 B
references/examples/go-api-with-react-admin/.scopesother93 B
references/examples/go-api-with-react-admin/AGENTS.mdmarkdown3.0 KB
references/examples/go-api-with-react-admin/admin/package.jsonmanifest423 B
references/examples/go-api-with-react-admin/admin/src/App.tsxother101 B
references/examples/go-api-with-react-admin/cmd/api/main.goscript75 B
references/examples/go-api-with-react-admin/go.modmanifest31 B
references/examples/go-api-with-react-admin/main.goscript71 B
references/examples/go-with-internal-web-tsx/.scopesother102 B
references/examples/go-with-internal-web-tsx/AGENTS.mdmarkdown3.1 KB
references/examples/go-with-internal-web-tsx/go.modmanifest34 B
references/examples/go-with-internal-web-tsx/internal/web/AGENTS.mdmarkdown3.1 KB
references/examples/go-with-internal-web-tsx/internal/web/package.jsonmanifest329 B
references/examples/go-with-internal-web-tsx/internal/web/src/App.tsxother57 B
references/examples/go-with-internal-web-tsx/internal/web/src/Button.tsxother63 B
references/examples/go-with-internal-web-tsx/internal/web/src/Footer.tsxother63 B
references/examples/go-with-internal-web-tsx/internal/web/src/Header.tsxother63 B
references/examples/go-with-internal-web-tsx/internal/web/src/Sidebar.tsxother65 B
references/examples/go-with-internal-web-tsx/main.goscript85 B
references/examples/go-with-internal-web-tsx/package-lock.jsonmanifest87 B
references/examples/go-with-internal-web-tsx/package.jsonmanifest221 B
references/examples/ldap-selfservice/AGENTS.mdmarkdown2.5 KB
references/examples/ldap-selfservice/go.modmanifest45 B
references/examples/ldap-selfservice/internal-AGENTS.mdmarkdown10.0 KB
references/examples/ldap-selfservice/internal-web-AGENTS.mdmarkdown12.1 KB
references/examples/php-with-frontend/.scopesother89 B
references/examples/php-with-frontend/AGENTS.mdmarkdown3.1 KB
references/examples/php-with-frontend/composer.jsonconfig185 B
references/examples/php-with-frontend/package.jsonmanifest307 B
references/examples/php-with-frontend/pnpm-lock.yamlmanifest101 B
references/examples/php-with-frontend/src/Controller.phpscript41 B
references/examples/php-with-frontend/web/AGENTS.mdmarkdown3.1 KB
references/examples/php-with-frontend/web/package.jsonmanifest515 B
references/examples/php-with-frontend/web/src/App.tsxother63 B
references/examples/php-with-frontend/web/src/Button.tsxother215 B
references/examples/php-with-frontend/web/src/Footer.tsxother125 B
references/examples/php-with-frontend/web/src/Header.tsxother122 B
references/examples/php-with-frontend/web/src/main.tsxother41 B
references/examples/php-with-frontend/web/tsconfig.jsonconfig280 B
references/examples/pnpm-workspace/AGENTS.mdmarkdown2.6 KB
references/examples/pnpm-workspace/package.jsonmanifest105 B
references/examples/pnpm-workspace/packages/web/package.jsonmanifest187 B
references/examples/pnpm-workspace/packages/web/src/index.tsxother258 B
references/examples/pnpm-workspace/pnpm-lock.yamlmanifest920 B
references/examples/pnpm-workspace/pnpm-workspace.yamlconfig27 B
references/examples/simple-ldap-go/AGENTS.mdmarkdown2.5 KB
references/examples/simple-ldap-go/examples-AGENTS.mdmarkdown1.8 KB
references/examples/simple-ldap-go/go.modmanifest43 B
references/examples/t3x-rte-ckeditor-image/AGENTS.mdmarkdown2.5 KB
references/examples/t3x-rte-ckeditor-image/Classes-AGENTS.mdmarkdown12.0 KB
references/examples/t3x-rte-ckeditor-image/composer.jsonconfig174 B
references/feedback-memory-schema.mdmarkdown6.2 KB
references/fleet-sync-sweep.mdmarkdown4.8 KB
references/git-hooks-setup.mdmarkdown2.5 KB
references/output-structure.mdmarkdown7.1 KB
references/quality-rubric.mdmarkdown6.2 KB
references/scripts-guide.mdmarkdown7.7 KB
references/verification-guide.mdmarkdown9.4 KB
scripts/analyze-git-history.shexecutablescript10.2 KB
scripts/check-freshness.shexecutablescript8.9 KB
scripts/detect-golden-samples.shexecutablescript5.7 KB
scripts/detect-heuristics.shexecutablescript3.2 KB
scripts/detect-project.shexecutablescript19.6 KB
scripts/detect-scopes.shexecutablescript13.4 KB
scripts/detect-utilities.shexecutablescript4.6 KB
scripts/extract-adrs.shexecutablescript6.8 KB
scripts/extract-agent-configs.shexecutablescript9.3 KB
scripts/extract-architecture-rules.shexecutablescript19.9 KB
scripts/extract-ci-commands.shexecutablescript12.5 KB
scripts/extract-ci-rules.shexecutablescript16.1 KB
scripts/extract-commands.shexecutablescript11.9 KB
scripts/extract-documentation.shexecutablescript7.8 KB
scripts/extract-github-rulesets.shexecutablescript3.8 KB
scripts/extract-github-settings.shexecutablescript3.7 KB
scripts/extract-ide-settings.shexecutablescript7.0 KB
scripts/extract-platform-files.shexecutablescript8.0 KB
scripts/extract-quality-configs.shexecutablescript14.9 KB
scripts/generate-agents.shexecutablescript118.2 KB
scripts/generate-file-map.shexecutablescript4.6 KB
scripts/lib/config-root.shexecutablescript5.8 KB
scripts/lib/git.shscript1.2 KB
scripts/lib/summary.shscript11.7 KB
scripts/lib/template.shexecutablescript15.9 KB
scripts/score-agents.shexecutablescript9.8 KB
scripts/tests/test-agents-file-write-boundary.shscript5.8 KB
scripts/tests/test-agents-scan-dependency-trees.shexecutablescript4.5 KB
scripts/tests/test-changelog-rule.shexecutablescript3.4 KB
scripts/tests/test-claude-import-file.shscript3.3 KB
scripts/tests/test-command-allowlist.shexecutablescript4.7 KB
scripts/tests/test-dependency-tree-exclusion.shscript2.5 KB
scripts/tests/test-extract-commands-real-scripts.shexecutablescript4.0 KB
scripts/tests/test-extract-github-settings.shexecutablescript2.0 KB
scripts/tests/test-git-history-counts.shexecutablescript3.6 KB
scripts/tests/test-golden-sample-file-names.shexecutablescript2.2 KB
scripts/tests/test-json-manifest.shexecutablescript5.5 KB
scripts/tests/test-last-verified-opt-in.shscript1.6 KB
scripts/tests/test-make-target-lookup.shscript2.5 KB
scripts/tests/test-merge-heuristic.shscript1.8 KB
scripts/tests/test-scope-index-heading.shexecutablescript4.4 KB
scripts/tests/test-scope-index-keeps-existing.shexecutablescript2.5 KB
scripts/tests/test-smoke-clock-step.shexecutablescript1.7 KB
scripts/tests/test-symlink-write-boundary.shexecutablescript6.3 KB
scripts/tests/test-untrusted-git-config.shscript5.0 KB
scripts/tests/test-verification-sidecar.shscript2.7 KB
scripts/validate-structure.shexecutablescript13.3 KB
scripts/verify-commands.shexecutablescript32.7 KB
scripts/verify-content.shexecutablescript13.2 KB

references/quality-rubric.md

View on GitHub
<!-- SPDX-License-Identifier: CC-BY-SA-4.0 -->
<!-- SPDX-FileCopyrightText: Netresearch DTT GmbH -->

# AGENTS.md Quality Rubric

Grade AGENTS.md files so you know which ones most need work. The grade has two
layers that are **kept separate on purpose**:

1. **Deterministic score** — `${CLAUDE_SKILL_DIR}/scripts/score-agents.sh PATH`. A 0-100 from the four
   verifier scripts; no model call, so re-running on an unchanged tree gives the
   same grade — except a file dated *today*, where git's bare-date `--since` can
   shift the Currency axis within the day.
2. **Qualitative LLM overlay** — three axes a *reading agent* judges (below).
   These vary between runs, so they are **never folded into the deterministic
   number**; present them as a separate annotation.

## Deterministic axes (score-agents.sh)

| Axis | Max | Source | What it measures |
|------|-----|--------|------------------|
| Structure | 25 | `validate-structure.sh --json` | Managed header, thin root, precedence, scope links, CLAUDE.md symlink, required scoped sections |
| Currency | 20 | `check-freshness.sh --json` | Commits in scope since the file's "Last updated" date |
| Content | 20 | `verify-content.sh --json` | Documented files/commands/counts that match reality |
| Commands | 15 | `verify-commands.sh --json` | Documented commands that actually run (root only) |
| Conciseness | 20 | line count vs budget | Root ≤50 lines ideal; scoped files get a larger budget |

Per file: `percent = earned / (sum of applicable axis maxima) × 100`. Scoped
files have no Commands axis (maxima sum to 85, normalised to 100).

Grades: **A ≥90 · B ≥75 · C ≥50 · D ≥30 · F <30**.

## Qualitative LLM overlay (agent-judged)

Run after the deterministic score. Read each AGENTS.md and rate three axes the
scripts cannot measure. Use **strong / adequate / weak** with a one-line reason.

| Axis | Strong | Weak |
|------|--------|------|
| **Architecture clarity** | A new agent can place code from the file alone — key dirs, module relationships, entry points | Vague or missing structure; "see the code" |
| **Actionability** | Concrete, copy-paste commands, real paths, decisive heuristics ("squash-merge", "ask first") | Theoretical advice ("follow best practices", "write good tests") |
| **Non-obvious patterns** | Captures what code can't tell you — ordering deps, quirks, "why we do it this way" | Only restates what the filenames already say |

### A convention needs a fill-in skeleton, not a paragraph

Actionability is not only about commands. Where the file states a *convention*
the agent has to reproduce — commit trailers, a message format, a required
header — prose is followed partially even when it is read, because the agent's
own defaults fill the gap. Give it a skeleton to copy, near the top.

Measured 2026-09-18 on `TYPO3-Documentation/TYPO3CMS-Reference-CoreApi`, whose
`AGENTS.md` states its commit trailers as prose in two numbered rules: a clone
with one uncommitted edit, a headless session asked only for the commit
message, Haiku 4.5, six runs per variant. The unchanged file is the baseline
and was run twice, once against each later batch, so its column pools 12
observations against the variants' 6.

| | prose rules | + skeleton | + skeleton and "replace the example" |
|---|---|---|---|
| required trailer present and correct | 4 / 12 | 6 / 6 | 6 / 6 |
| documented `Assisted-by:` form | 1 / 12 | 6 / 6 | 6 / 6 |
| model correctly named | 0 / 12 | 1 / 6 | 6 / 6 |

Two things the last column pays for:

- **A concrete example is copied verbatim.** Five of six runs filed
  `Assisted-by: Claude Sonnet 5` while a different model was running — a
  skeleton turns a placeholder into an assertion unless it says to replace it.
- **A placeholder loses the field.** Writing `<model name> <contact>` instead
  removed the false attribution by removing the trailer: it then appeared in
  1 of 6. Keep the example *and* say to replace it.

The control matters for the axis, too. Removing the one-line `CLAUDE.md` that
holds `@AGENTS.md`, and leaving `AGENTS.md` itself in place, took every
trailer to 0 of 6 — the same as deleting both, 0 of 4. So in the runs above
the rules were reaching the session the whole time, and what the skeleton
fixes is compliance, not delivery. "Not followed" and "not loaded" look
identical in the output and have different fixes; verify which one you have
before rewriting content. Claude Code reaches the rules only through
`CLAUDE.md`, never by the AGENTS.md name alone; see
[`ai-tool-compatibility.md`](ai-tool-compatibility.md).

### How to feed it back: `--review`

Rate each file, write the ratings to JSON, and pass it to the scorer. It keeps the
deterministic grade as the headline and adds a clearly-labelled, **non-reproducible**
secondary grade:

```bash
cat > /tmp/review.json <<'JSON'
{
  "AGENTS.md":     {"architecture":"strong","actionability":"adequate","non_obvious":"weak"},
  "web/AGENTS.md": {"architecture":"weak","actionability":"weak","non_obvious":"weak"}
}
JSON
bash ${CLAUDE_SKILL_DIR}/scripts/score-agents.sh PATH --review /tmp/review.json
```

```
  B  78/100    AGENTS.md
       structure 25/25  currency 15/20  content 18/20  conciseness 20/20  commands 0/15
       with review: B ~75/100  (non-reproducible)
```

Rating → fraction of the axis max: **strong = 1.0 · adequate = 0.6 · weak = 0.2**.
Blended = `(deterministic_earned + llm_earned) / (deterministic_max + 25) × 100`.
The headline `percent`/`grade` are **byte-identical** with or without `--review` —
the overlay never moves the reproducible number.

- The overlay's job is to catch files that **score well but read poorly** — e.g.
  structurally complete (B+) yet every line is obvious, or actionability is vague.
  Those are the highest-value edits.
- The blended figure varies between reviews (model judgement); the deterministic
  headline stays put. Always cite both, never the blended number alone.

## Workflow

1. `${CLAUDE_SKILL_DIR}/scripts/score-agents.sh PATH` → deterministic report (worst-first).
2. Open the worst-graded files; run the three overlay axes on each.
3. Fix highest-leverage gaps first (a `D` on Structure is mechanical; a `weak`
   on Non-obvious patterns needs human/session knowledge — see
   [`feedback-memory-schema.md`](feedback-memory-schema.md) and the retro-skill).
rea-changelog-updatemorluto/reaPrepare or rewrite REA release changelogs and GitHub release notes from pinned Git history, with verified contributor thanks and Release Please synchronization.Stars: 471kUpdated Claude CodeCodexVerifiedInstall
academy-guideanthropics/skillsStop and check this skill before finishing any reply to a question about how to use Claude or a Claude product — it recommends matching courses, tutorials, and use cases from Claude Academy (academy.claude.com), Anthropic's learning hub. Trigger on: "how do I", "how can I", "getting started with", "what can Claude do", "teach me", "learn to use"; questions about artifacts, projects, skills, plugins, connectors, MCP; requests about rolling Claude out to a team, class, or organization; and any ask for training materials, onboarding content, or learning resources. Use it when the user is learning how to use a feature or product — not when they are mid-task and just want the task done. This skill composes with other skills: after consulting product documentation to answer how a Claude feature works, also check here for a matching course or tutorial — a docs-grounded answer and an Academy recommendation belong together. Only recommend on a strong match; never invent Academy content.Stars: 198kUpdated Claude CodeCodexVerifiedInstall
claude-apianthropics/skillsReference for the Claude API / Anthropic SDK — model ids, pricing, params, streaming, tool use, MCP, agents, caching, token counting, model migration. TRIGGER — read BEFORE opening the target file; don't skip because it "looks like a one-liner" — whenever: the prompt names Claude/Anthropic in any form (Claude, Anthropic, Fable, Opus, Sonnet, Haiku, `anthropic`, `@anthropic-ai`, `claude-*`, `us.anthropic.*`, `[1m]`); the user asks about an LLM (pricing/model choice/limits/caching) — never answer from memory; OR the task is LLM-shaped with provider unstated (agent/MCP/tool-definition/multi-agent/RAG/LLM-judge/computer-use; generate/summarize/extract/classify/rewrite/converse over NL; debugging refusals/cutoffs/streaming/tool-calls/tokens). SKIP only when another provider is being worked on (overrides all triggers): OpenAI/GPT/Gemini/Llama/Mistral/Cohere/Ollama named in the query; OR `grep -rE 'openai|langchain_openai|google.generativeai|genai|mistralai|cohere|ollama'` over the project hits (run this grep FIRST if no provider named — don't Read the file).Stars: 198kUpdated Claude CodeCodexUnverifiedInstall
doc-coauthoringanthropics/skillsGuide users through a structured workflow for co-authoring documentation. Use when user wants to write documentation, proposals, technical specs, decision docs, or similar structured content. This workflow helps users efficiently transfer context, refine content through iteration, and verify the doc works for readers. Trigger when user mentions writing docs, creating proposals, drafting specs, or similar documentation tasks.Stars: 198kUpdated Claude CodeCodexVerifiedInstall