agentleFS
formal-youformal-you/agent-skills-suite/skills/workflow/review/SKILL.md

review

Skill0 starsChanged 28 days agoMITView on GitHub

Perform an independent, read-only review of a plan/task bundle (Gate 2) or a code change, producing evidence-backed findings and one verdict without editing anything. Use when Gate 1 has passed and the plan needs semantic review, or when a change needs review. Do not use for implementation, planning, QA execution, or reviewing your own work.

192 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use close-spec, design-with-images and 10 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • dzhngSkill

    close-spec

    When a spec under specs/<feature>/ is done shipping (a write-spec build, or any planned task that produced a spec), archive it to specs/done/ and rewrite it from a build-plan into a durable rationale — the why, the principles, the invariants — pointing back to the real code for the how. Use when implementation has landed and the plan no longer matches what shipped, or the user says a feature/spec is finished. Pairs with [write-spec](../write-spec/SKILL.md) (the plan this closes) and [review](../review/SKILL.md) (run before closing).

    Also in 5 projects that use this

  • dzhngSkill

    design-with-images

    Explore UI and visual design with image generation, then iterate the real implementation against the user-selected concept. Use when brainstorming visual directions, offering mockup variants, or implementing an approved generated design; do not stop at the mockups or the first code approximation.

    Also in 5 projects that use this

  • LambenthanSkill

    daily-arxiv

    每日从 arXiv 拉取新论文,过滤相关性,auto-ingest 高优先级论文,检测 SOTA 更新

    Also in 5 projects that use this

  • LambenthanSkill

    exp-eval

    实验判决门:Review LLM 独立评判实验结果 → 4 种判决路径 → 自动更新 claims confidence、ideas status、graph edges

    Also in 5 projects that use this

  • +3 more from dzhng
  • +5 more from Lambenthan

The file

We split this skill into the rules it gives an agent

8 rules. 0 are common, 4 are distinctive, and 4 are very specific to this project.

  • 0

    Common

    Most of the 157 edited versions we read keep these.

  • 4

    Distinctive

    This author's own choices. Worth a look.

  • 4

    Very specific

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

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

    Test-design quality review belongs to the independent test-engineering review mode.

    line 32
  2. Very specific:

    Use this mode only when a Dev files an Appeal Request against a REWORK_REQUIRED verdict.

    line 46
  3. Very specific:

    Every tasks/<task-id>/01-contract.md and the Gate 1 result.

    line 59
  4. Distinctive:

    No edits to any artifact. Review notes go in the review response, not in the task documents.

    line 71
  5. Distinctive:

    Completeness: every requirement, state transition, API, entity and invariant maps to a task.

    line 95
  6. Very specific:

    Zero hallucination: every task entity/field/endpoint/error code traces to upstream text.

    line 96
  7. Distinctive:

    ADR compliance: no task contradicts an inherited ADR.

    line 98
  8. Distinctive:

    Security and safety: no secrets, injection, unsafe deserialization, missing authorization or unintended data exposure.

    line 109

“Common” is measured: kept by at least 6 in 10 of the 157 edited versions of this file we could read. “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 — Independent, Read-Only Review

Review the work of another role, find what is wrong or missing, and issue a verdict. This skill defines the review process; it never edits the artifact under review and never replaces Gate 1 or QA verification.

Use when

  • Gate 2 (plan review): Gate 1 passed and the task bundles need a semantic check against the PRD, ADRs and tech spec before implementation.
  • Change review: a Dev has implemented a task and the change needs a quality/contract review (optionally before or alongside QA verification).
  • A reviewer is explicitly asked to audit traceability, scope or architecture compliance.

Do not use when

  • You wrote the artifact under review — independence is required; escalate to another reviewer.
  • Gate 1 is failing or the plan is still being written — fix the plan first.
  • You are asked to implement, plan or run acceptance verification — those belong to Dev, Planner and QA.
  • The review would require editing the artifact to make it pass. Report findings instead.
  • Test-design quality review belongs to the independent test-engineering review mode; this skill stays semantic plan/change review.

Review modes

ModeReview objectPrimary question
Plan / Gate 2.specs/ + tasks/<task-id>/01-contract.mdIs the plan truthful, complete and ADR-compliant?
Change / codediff/commit + contract + test evidenceDoes the change satisfy the contract without scope, safety or quality regressions?

Choose one mode per review; do not mix a plan review with code review in one verdict.

Arbitration mode

Use this mode only when a Dev files an Appeal Request against a REWORK_REQUIRED verdict. Work as the independent third layer, not as a second QA and not as the Dev's helper: arbitrate from the contract, the spec anchor, QA failure evidence and the appeal, and return exactly one decision — CODE_BUG (to Dev), TEST_BUG (to QA) or SPEC_AMBIGUITY (to the Planner/human) — with evidence line numbers. Read references/arbitration.md for inputs, independence and the decision table. Arbitration never edits any file.

Inputs

Plan review

  • PRD (.specs/01-prd.md / .specs/01-prd/), ADRs (.specs/adr/), tech spec (.specs/02-tech-spec.md / .specs/02-tech-spec/).
  • Every tasks/<task-id>/01-contract.md and the Gate 1 result.

Change review

  • The contract (01), the diff/commit, 02-dev-workspace.md evidence, and the acceptance/test result.
  • The repository standards and ADRs the change must honor.

Outputs

  • A findings list with severity, evidence (file/line or diff hunk) and the required action.
  • Exactly one verdict: PASS, REQUEST CHANGES or BLOCKED.
  • No edits to any artifact. Review notes go in the review response, not in the task documents.

Workflow

  1. Check independence. Confirm you did not author the artifact. If you did, stop and hand the review to another reviewer.
  2. Pick the mode. Plan review runs after Gate 1. Change review runs on a concrete diff and its evidence; do not review an unbuilt plan as code.
  3. Read upstream first. For plan review, read PRD/ADR/spec before the contracts. For change review, read the contract and the diff before judging style.
  4. Apply the checklist. Plan review: completeness, zero hallucination, ADR compliance, contract quality. Change review: contract alignment, scope, correctness, tests/edge cases, security, maintainability. Use references/plan-review.md and references/code-review.md.
  5. Classify findings. blocker (cannot proceed), major (must fix before merge/handoff), minor (should fix), nit (optional). Every finding needs evidence and a concrete action.
  6. Issue one verdict. PASS only when the checklist was actually exercised and no blocker/major remains. REQUEST CHANGES for fixable findings. BLOCKED when inputs are missing or the reviewer cannot be independent.
  7. Hand off. Plan findings go to the Planner; change findings go to Dev (and QA if verification is affected). Do not edit on their behalf.

Plan review essentials (Gate 2)

  • Completeness: every requirement, state transition, API, entity and invariant maps to a task.
  • Zero hallucination: every task entity/field/endpoint/error code traces to upstream text; no template examples (cart, order, payment, coupon, tenant_id, etc.) unless the PRD defines them.
  • ADR compliance: no task contradicts an inherited ADR.
  • Contract quality: task_id matches the directory, touch_files is 1–3, DoD commands and edge cases are concrete, dependencies are acyclic.
  • Gate 1 must pass before Gate 2; never sign off a plan that Gate 1 rejected.

Change review essentials

  • Contract alignment: the diff implements what 01 declares; every changed path is in touch_files; no scope creep.
  • Correctness and edge cases: behavior matches the contract; the declared edge cases are handled.
  • Tests/evidence: tests exist and pass; the evidence in 02 is real (command + result), not a claim.
  • Security and safety: no secrets, injection, unsafe deserialization, missing authorization or unintended data exposure.
  • Maintainability: follows repository standards; avoid stylistic nits unless the repo enforces them.

Must not

  • Edit the artifact, the contract, tests, source or 04-process.md.
  • Weaken, skip or bypass Gate 1, or sign off while Gate 1 is failing.
  • Rewrite a contract to make it pass, or accept "done" without evidence.
  • Review your own work or issue a verdict after only skimming.

Handoff

  • Plan review: PASS allows implementation; REQUEST CHANGES returns the task set to the Planner.
  • Change review: PASS allows merge/QA handoff; REQUEST CHANGES returns the diff to Dev.
  • BLOCKED means missing inputs or independence; state exactly what is missing.

References

  • references/plan-review.md — Gate 2 checklist and output format.
  • references/code-review.md — change/code review checklist.
  • references/arbitration.md — appeal arbitration: inputs, independence rules and the CODE_BUG | TEST_BUG | SPEC_AMBIGUITY decision table.
  • references/role-boundaries.md — read-only invariants and the honest enforcement model (bundled; no external files needed).
  • Planning gate: the plan-linter skill when installed; its Gate 2 fallback checklist is also bundled here as references/plan-review.md.
  • Repository rules: read AGENTS.md and any role-rules files when the adopting project defines them; they may add requirements but are not required for this skill.

Honest limits

Semantic review is prompt-driven, not automated. It can miss issues and cannot guarantee security. Gate 1 remains the deterministic structural gate; QA remains the independent execution/verification gate. This skill provides a third, read-only judgment layer.

Image: untitled

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.