agentleFS
kapadiaskapadias/nonna/.claude/skills/review/SKILL.md

review

Skill0 starsChanged 47 days agoMITView on GitHub

Proportional review before merge — review-lanes.sh sizes it: one cheap reviewer for a fast-lane diff, full review otherwise, plus a security review when a risky path or line is touched.

192 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use close-spec, daily-arxiv 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

  • 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

  • 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

    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

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

The file

We split this skill into the rules it gives an agent

3 rules. 0 are common, 1 is distinctive, and 2 are very specific to this project.

  • 0

    Common

    Most of the 156 edited versions we read keep these.

  • 1

    Distinctive

    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:

    Size it — the script decides: Run bash "${CLAUDE_SKILL_DIR}/scripts/review-lanes.sh".

    line 17
  2. Distinctive:

    Persist each verdict — verbatim: Every reviewer ends with exactly one fenced json verdict block.

    line 28
  3. Very specific:

    Run the gate — the script decides: Run bash "${CLAUDE_SKILL_DIR}/../code-review/scripts/check-review.sh" on each verdict file.

    line 33

“Common” is measured: kept by at least 6 in 10 of the 156 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

!git branch --show-current !git status --short !git diff develop...HEAD --stat

Review: $ARGUMENTS (if empty, review the current branch's diff vs develop).

Steps

  1. Size it — the script decides. Run bash "${CLAUDE_SKILL_DIR}/scripts/review-lanes.sh". It prints lane=light|full and security=yes|no, and fails closed to full / yes. Anything but exactly lane=light means full; anything but exactly security=no means yes — including a run that fails or prints nothing. Never re-classify by judgement, up or down (ADR-0009).
  2. Launch reviewers in PARALLEL (independent — do not serialize):
    • Always: code-reviewer — correctness, broken contracts, missing/weak tests, silent failures, reproducibility, style fit. On lane=light, launch it with model: sonnet: a diff the fast-lane classifier accepts does not need the deep tier.
    • If security=yes: also security-reviewer, at its own tier — injection, secret leakage, broken authz, unsafe deserialization, supply chain.
  3. Persist each verdict — verbatim. Every reviewer ends with exactly one fenced json verdict block. Write each block byte-for-byte — no merging, no rewriting, no "cleanup" — to .claude/reviews/<sha>-code.json and (when the security reviewer ran) .claude/reviews/<sha>-security.json, where <sha> is git rev-parse --short HEAD. These are transient gate inputs, git-ignored; a new commit invalidates them by construction.
  4. Run the gate — the script decides. Run bash "${CLAUDE_SKILL_DIR}/../code-review/scripts/check-review.sh" on each verdict file. A non-zero exit means the review gate is red. Report the script's output as the verdict and never override it — the parser, not the model, decides merge-readiness (ADR-0005).
  5. Debt gate. Run bash "${CLAUDE_SKILL_DIR}/../lean/scripts/check-debt.sh" --range develop...HEAD (the same range as the diff). A new debt: marker with no upgrade trigger is a gate failure (exit 1) — report it alongside the verdict gates; it is fixed by naming the trigger, never by deleting the comment while keeping the corner.
  6. Synthesize for the human. Merge findings into one report, deduplicated, grouped by severity — CRITICAL / HIGH / MEDIUM / LOW — each with path:line, the issue, and a concrete fix. Any breach of a trust boundary, a safety gate, or "fail-closed" is automatically CRITICAL. The prose explains; the gate's exit code decides. If any finding carries category: simplicity, end with the line net: -N lines possible.

Output

The gate result (per verdict file), then the consolidated, severity-grouped findings. If any issues were fixed inline, note them and re-run /review — verdicts are per-commit, and the lanes are re-computed on the new diff. Update the tracked issue with the outcome.

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.