repo-review
Skill10 starsChanged 32 days agoMITView on GitHub
Use when reviewing a pull request or a change in this repository — orients the reviewer to this repository's standards (AGENTS.md conventions, decision log, gates) and the review-specific checks that code alone cannot show
2 other projects use a version of this.
Adopted by
Most recently taken up by
- idaibin/skillstheir own version, first seen Nov 8, 2026
- taoquo/three-skillstheir own version, first seen Apr 1, 2026
Other versions2 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 2
- idaibin/skills★ 6same name
- taoquo/three-skills★ 4same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository6 other files it gives its agents.
- AGENTS.mdAGENTS.md
- docs/AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- repo-decisionsSkill
- repo-governanceSkill
- repo-seedSkill
The file
We split this skill into the rules it gives an agent
1 rule. About 0 are common, 1 is distinctive, and 0 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.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 34
Capability authority stays explicit: Read-only audits and dry-runs may be autonomous.
≈ “Common” is an estimate: at least 1 in 5 of the 372,429 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
Reviewing a change in this repository
Read the diff, the owning docs, the decision log, and enough surrounding code to understand the design before judging it. Blocking requirements are hard: a violation blocks the change. Manual checks rank remaining risk by this project's real failure modes — apply the ones the change touches, not every one, to every change.
Sources of truth
- AGENTS.md: standing repository rules.
- docs/AGENTS.md: documentation placement and prose discipline.
- docs/decisions/: durable design rationale. Disagreement with a decision record is a design discussion, not an automatic veto.
- docs/specs/: risk-triggered change contracts.
- docs/testing.md: risk-to-layer selection, test topology, evidence rules, and maintenance budget.
- docs/architecture.md: the module map and seams.
Blocking requirements
Universal (applies to any repository)
- New prose receives semantic review. Critically review every added or changed Markdown passage, JSDoc, comment, prompt, description, diagnostic, and visible string. Verify required coverage, accuracy, placement, and editorial quality against the owning code or behavior; automated checks do not establish those properties.
- Docs match the code. Config, defaults, errors, wire fields, events, and public behavior update the owning docs and JSDoc in the same diff.
- Contracts and decisions use the right artifact. A risk-boundary change has an Approved spec; a durable choice with real alternatives has a decision record. Do not demand an ADR as a change log.
- Tests provide minimum sufficient behavior evidence. Name the meaningful regression each changed test prevents and place its primary evidence at the lowest sufficiently real boundary. A fix links an existing deterministic reproduction or adds a regression test that fails before the fix and passes afterward; redundant coverage or implementation-only assertions do not satisfy this requirement.
- External-source provenance is retained. If implementation is materially derived from a paper, article, community post, benchmark, research report, or copied/adapted code, cite it at the closest stable code location or link that location to a decision record whose
## Linkscites the source. A pull request, issue, prompt, or chat-only citation does not count; copied or adapted material also preserves applicable license and NOTICE requirements.
Project-specific (instantiated at seed time from repo-seed's own rules)
- Seeded surface stays bounded. A change writes only the seeded paths;
references/templates/remains the single source of truth forseededFiles(), and every seeded-file change updates templates, manifest, and tests in the same diff. - Update semantics are preserved. Changes to
scripts/scaffold.mjskeepreferences/update-strategy.mdand the seeded.repo-seed/update-strategy.mdin sync; every preservation rule (never overwrite user-modified or instantiated content, never delete user files, never regress instantiated content to placeholders) has a test. - Tokens ship with instantiation guidance. Any new template token needs coverage in
SKILL.mdstep 4 orreferences/review-standard.md, plus a refusal-path test. - Capability authority stays explicit. Read-only audits and dry-runs may be autonomous; enabling capabilities, installing hooks, changing governance policy/source-of-truth, or connecting external systems requires recorded user authority.
- Resident governance routing stays explicit. Root AGENTS identifies the repo-seed-managed governance boundary and routes review, durable decisions, governance evolution, and lifecycle upgrades to the correct resident or global skill without making repo-seed a dependency of ordinary implementation.
Manual checks
Project-specific (instantiated at seed time from repo-seed's stack and known failure modes)
- Plan/apply symmetry:
planRundecisions (create/update/skip/refusal) andapplyPlanmanifest writes stay consistent; dry-run never mutates disk or manifest. - Gate determinism: verifier scopes never silently shrink; a verifier exits non-zero only on real violations.
- Capability lifecycle: recommendation hashes depend only on normalized relevant facts; declined/deferred suggestions stay suppressed until those facts or rules change; enabled capabilities remain managed when CLI flags are omitted.
- Adoption uniqueness: non-standard paths and external sources of truth are registered and linked, not copied into a competing governance system.
- Template ↔ dogfood parity (static files): static seeded files in this repository match their templates; the instantiated repo-review and manifest hashes are exempt by design.
- Refusal correctness: a re-run that would regress instantiated content to unresolved placeholders is refused;
userModifiedentries are never refreshed. - Model perspective: repo-seed is agent-facing — review changes to
SKILL.md, templates, and seeded prose as instructions a fresh agent will read: check the exact wording and that tokens, links, and paths resolve from a clean checkout.
Universal fallbacks (apply where the project has no specific rule)
- Intent and interface contracts: trace both sides of every changed interface. Confirm the implementation matches the change and any decision record, including errors, cancellation, ownership, and disposal.
- Lifecycle and concurrency: for async setup, callbacks, processes, or teardown, check races before publication, cancellation during awaits, independent error reporting, and complete cleanup.
- Human readability: read changed code from its entry point downward. The primary path stays understandable without opening every helper. Challenge abstractions that only forward or rename, mixed abstraction levels, misleading or generic names, new synonyms for existing concepts, and logic outside its owning boundary. Block when these make behavior unreliable to infer, hide effects or cost, or violate established vocabulary or boundaries; otherwise report subjective polish as advisory. Line counts, parameter counts, duplication, and complexity scores are signals, not verdicts.
- Scope and necessity: map each abstraction, option, defensive copy, and compatibility path to its current contract and consumer. Challenge unrelated features and speculative generality.
- Bounds cover the final operation: probe tiny and exact limits, oversized chunks, and multibyte text for byte limits.
- Test value and boundary: identify the contract each changed test owns, challenge duplicate evidence, and use the real shipped entry path whenever packaging, configuration, process setup, or wiring is part of the risk.
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.

