agentleFS
arxdsilvaarxdsilva/ai-loops/skills/review-integration/SKILL.md

review-integration

Skill0 starsChanged 7 days agoMITView on GitHub

Use when reviewing a plan, a diff or a wave of agent work against architecture, consistency, performance and test standards - plan-reviewer, qa-verifier and product-reviewer procedure, reject list and multi-agent integration.

2 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

3 rules. About 0 are common, 1 is distinctive, and 2 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.

  • 2

    Very specific

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

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

    Read only the acceptance criteria, the diff, the affected package AGENTS.md, the reported tests and benchmarks.

    line 10
  2. Very specific:

    Use templates/PLAN_REVIEW.md.

    line 47
  3. Distinctive:

    Before accepting a wave: collect claims reports.

    line 55

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 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.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Review and integration

Load

Read only the acceptance criteria, the diff, the affected package AGENTS.md, the reported tests and benchmarks. Definition of done: specs/WORKFLOW.md §8. Open more source only when needed (specs/CONTEXT_DISCIPLINE.md §19).

Review order

  1. correctness
  2. architecture ownership (specs/ARCHITECTURE.md §29)
  3. consistency: clock, RNG, ordering, idempotency (skills/consistency)
  4. invariants
  5. data complexity and performance
  6. concurrency and the pack's race policy (WORKFLOW §6)
  7. persistence, API and config compatibility
  8. tests: behaviour named, failure paths, no weakened tests
  9. API and readability per the language's idioms (skills/lang-<id>/idioms.md; Go: specs/languages/go/RULES.md §51)
  10. documentation: package AGENTS.md for changed contracts
  11. observability: errors explicit, logs and metrics where operators need them

Reject

  • an unrelated broad refactor;
  • a weakened or deleted test;
  • a duplicate canonical type;
  • domain depending on an adapter, transport or storage type;
  • global mutable state, import-time registration (Go init());
  • a thread, task or goroutine per item without a bound;
  • the wall clock or a global RNG in authoritative code (the pack's impure_calls);
  • decisions from hash-map iteration order;
  • a silent persisted-schema or API break;
  • a retried operation without an idempotency key;
  • a new hot global scan or N+1 query;
  • a performance claim without a benchmark;
  • unbounded external input;
  • hand-edited generated code;
  • business rules in a handler or repository;
  • a claim without evidence, or evidence the reviewer cannot re-run.

Reviewing a plan (plan-reviewer)

Use templates/PLAN_REVIEW.md. Check completeness against specs/PARALLEL_WORK.md §1: contract, waves table, per-agent blocks, disjoint ownership, models with reasons, QA pairing, proof plan. Findings are numbered, each with what, why and a suggested change; only blockers block.

Verifying claims (qa-verifier)

Independently re-run the evidence, read the code, attack the claim with edge cases and the pack's race policy, and report per claim VERIFIED or FAILED with a repro (templates/QA_VERDICT.md). List unclaimed problems separately. Do not fix code.

Multi-agent integration

Before accepting a wave: collect claims reports; compare interface changes; find overlapping files; resolve each canonical contract once; merge in dependency order; run loops gate (every detected pack) and the affected benchmarks (specs/PARALLEL_WORK.md §6). Never keep two competing abstractions to "clean up later".

Report check

Each report states: changed contracts, invariants, tests, benchmark delta, consistency impact, compatibility impact, limits and follow-ups. Do not request chain-of-thought.

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.