repository-guidance-authoring
Skill1 starsChanged 16 days agoApache-2.0View on GitHub
Create or improve canonical agent guidance files (AGENTS.md, repository instructions, onboarding docs) generated from repository evidence. Use when asked to write or update a project's agent instructions. Do not use to author reusable skills (use skill-authoring), expose guidance to tools (use harness-adaptation), or check doc factual correctness (use documentation-review); this skill writes guidance, it does not build skills or verify prose.
In this repository25 other files it gives its agents.
- AGENTS.mdAGENTS.md
- adversarial-pr-reviewSkill
- ai-slop-detectorSkill
- architecture-reviewSkill
- codebase-orientationSkill
- code-reviewSkill
- dependency-upgradeSkill
- documentation-reviewSkill
- frontend-quality-reviewSkill
- git-github-workflowSkill
- harness-adaptationSkill
- implementation-planningSkill
- parallel-multi-agentSkill
- quality-hardeningSkill
- reduce-code-sizeSkill
- requirements-and-designSkill
- review-feedback-resolutionSkill
- rules-and-skills-auditSkill
- security-reviewSkill
- skill-authoringSkill
- skill-discoverySkill
- skill-optimizerSkill
- skill-reviewerSkill
- systematic-debuggingSkill
- threat-modelingSkill
The file
We split this skill into the rules it gives an agent
6 rules. About 0 are common, 5 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈5
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
- Distinctive: line 27
Use this skill when the deliverable is a guidance file itself, not a skill and not a doc-accuracy audit.
- Distinctive: line 50
When two applicable sources conflict and precedence cannot be established, show both claims and stop for a decision.
- Distinctive: line 65
When a policy is unknown, mark it as a placeholder for the team to confirm rather than asserting it.
- Distinctive: line 70
Keep claims verifiable; cite the file or configuration that supports each fact.
- Distinctive: line 71
Prefer a thin, accurate file over a long one padded with guesses.
- Very specific: line 72
Do not duplicate a skill's content.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill 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.
Files
Repository Guidance Authoring
Authority and boundary
This skill writes the canonical, project-local instructions an agent reads to
work in a repository — AGENTS.md, contribution/agent onboarding docs, and similar
instruction files. It is distinct from its neighbors:
| Skill | Owns |
| repository-guidance-authoring | Writing canonical agent guidance files from repository reality |
| skill-authoring | Authoring reusable, portable skills under skills/ |
| harness-adaptation | Exposing existing guidance to an agent harness's discovery |
| documentation-review | Checking a document's factual correctness against source truth |
Use this skill when the deliverable is a guidance file itself, not a skill and not a doc-accuracy audit.
Generate from evidence
Build the guidance from what the repository actually shows, not from a template of generic advice. Include sections as the evidence supports them:
- project purpose — what the software does, in user terms.
- setup commands — how to install and bootstrap, read from package/build files.
- development workflow — branching, building, and running locally.
- testing commands — how to run the relevant test suite.
- conventions — naming, layout, and contribution norms observed in the repo.
- build/deployment information — how the project builds and ships, when present.
- agent-specific rules — directives the project wants an agent to follow.
Scope and precedence
Inventory every applicable guidance file from the repository root to the target path before authoring. Put repository-wide rules at the root and narrower commands or constraints nearest the subtree that owns them. Nested guidance may refine its parent within that subtree, but it must not silently contradict higher-authority instructions. Do not restate system, user, or harness policy as repository policy. When two applicable sources conflict and precedence cannot be established, show both claims and stop for a decision.
Critical rule: facts versus policy
Separate what is discovered from what requires a human decision:
- FACTS — discovered from code, configuration, tests, and build behavior.
State these as observations. Example allowed: "The project uses pytest, based on
pyproject.toml." - POLICY — requires a human or team decision (merge strategy, review requirements, branch protection, coding mandates). Never invent team policy.
Example of a disallowed claim: "All PRs require squash merge" — only state this if repository evidence (a contributing guide, CI config, or branch rule) says so. When a policy is unknown, mark it as a placeholder for the team to confirm rather than asserting it.
Quality bar
- Keep claims verifiable; cite the file or configuration that supports each fact.
- Prefer a thin, accurate file over a long one padded with guesses.
- Do not duplicate a skill's content; if a reusable procedure belongs in a skill,
point to it rather than copying it here (per
harness-adaptation's single-source rule).
Avoid skill duplication
Repository guidance should contain project-specific rules and navigation, not copies
of reusable skill procedures. Prefer linking to skills over embedding their
workflows. A generated AGENTS.md or onboarding doc points at the canonical skill;
it does not become a second copy of it.
Output
Return the drafted guidance file (or the proposed edits to an existing one) with a
short note listing which statements are facts (and their evidence) and which are
policy placeholders awaiting team confirmation. Do not begin adopting the file into
a harness; that hand-off belongs to harness-adaptation.
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
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.

