agentleFS
ahonnahonn/dotfiles/.claude/skills/self-review/SKILL.md

self-review

Skill62 starsChanged 2 months agoMITView on GitHub

Review branch, staged, or uncommitted changes for actionable defects before commit, push, or PR creation.

52 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use latex-formatting, novelty-assessment 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.
  • lingzhi227Skill

    latex-formatting

    Handle LaTeX formatting, templates, and styling for academic papers. Set up conference templates (ICML, ICLR, NeurIPS, AAAI, ACL), fix formatting issues, manage packages, and ensure venue-specific compliance. Use when the user needs to set up a paper template, fix LaTeX formatting, or prepare for submission.

    Also in 4 projects that use this

  • lingzhi227Skill

    novelty-assessment

    Assess research idea novelty through systematic literature search. Multi-round search-evaluate loops with harsh critic persona. Binary novel/not-novel decision with justification. Use before committing to a research direction.

    Also in 4 projects that use this

  • michaelboedingSkill

    slide-generation

    Use this skill to create presentation slides from structured content. Triggers: "create slides", "generate presentation", "make powerpoint", "create pptx", "build slides", "presentation from content", "slide deck", "slides for" Outputs: PPTX files, Markdown slides, or HTML presentations. Used by: pitch-deck-agent, market-researcher-agent, and other agents needing slides.

    Also in 4 projects that use this

  • juanlurgSkill

    paper-to-code

    Reads a research paper and generates an implementation notebook with equations, code, and a toy example. Use when the user wants to implement a paper, convert a paper to code, or reproduce a paper's algorithm.

    Also in 4 projects that use this

  • OpenAISkill

    idea-generation

    Use when triaging public-equity idea candidates. Do not use for final trade recommendations, pitches, memos, or models.

    Also in 3 projects that use this

  • sunnyclinches44Skill

    explain

    Explains code using real-world analogies and ASCII diagrams

    Also in 5 projects that use this

  • +6 more from lingzhi227

The file

We split this skill into the rules it gives an agent

6 rules. 0 are common, 5 are distinctive, and 1 is very specific to this project.

  • 0

    Common

    Most of the 43 edited versions we read keep these.

  • 5

    Distinctive

    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
  1. Distinctive:

    Follow global AGENTS.md for authorization and test mode.

    line 12
  2. Distinctive:

    Read the full relevant diff and enough surrounding code to understand callers, dependencies, and error paths.

    line 31
  3. Very specific:

    Use an independent reviewer when --peer is requested, or when complex or high-risk changes need a second opinion.

    line 46
  4. Distinctive:

    Run local review alongside an independent review only when they can proceed independently.

    line 50
  5. Distinctive:

    Follow the global authorization boundaries for API or schema changes, destructive actions, and material scope changes.

    line 55
  6. Distinctive:

    After corrections, inspect the final diff and validate according to the global test mode and required project checks.

    line 59

“Common” is measured: kept by at least 6 in 10 of the 43 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

Self-Review

Review the requested diff against its intended behavior and surrounding contracts. Prioritize correctness, security, and maintainability. Use code-quality for relevant standards; avoid speculative findings or changes that only satisfy a preferred style.

Follow global AGENTS.md for authorization and test mode. PR creation warrants a local review, not an automatic independent-review requirement. Existing permission to implement or fix the work carries into review.

Arguments and scope

  • --base <branch>: Use the specified base. Otherwise inspect the PR base or repository default branch; verify the ref exists rather than guessing from a shell fallback.
  • --scope branch: Review committed changes relative to the merge base with the selected base.
  • --scope staged: Review git diff --cached.
  • --scope uncommitted: Review staged and unstaged changes, including relevant untracked files.
  • --fix: Apply technically verified fixes within the requested scope.
  • --report-only: Report findings without editing.
  • --peer: Request an independent review of the same scope using available reviewer tools.
  • Without a fix flag, preserve the session's authorization. A review-only request produces findings; an implementation or fix request permits relevant corrections.

Inspect git status, the current branch, and available refs before selecting scope. Without an explicit scope, cover the requested work's final state: include relevant branch commits and staged, unstaged, or untracked changes that belong to the task. Do not omit a newly implemented fix because earlier work is already committed. Keep unrelated user changes outside the review. An explicitly requested revision or scope limits the review to that selection.

Use the remote-tracking base when current enough for the task; fetch when freshness matters. For committed branch changes, use git diff <base>...HEAD and inspect the associated commits. When the default task scope also includes local work, inspect git diff HEAD and relevant untracked files, then read the combined final implementation. If the requested scope is empty or the base cannot be established, report that limitation rather than silently reviewing another scope.

Inspect the change

Read the full relevant diff and enough surrounding code to understand callers, dependencies, and error paths. Compare implementation with the requested outcome and, where relevant, the PR description or commit messages.

Prioritize checks according to the change:

  • Correctness: invariants, edge cases, nullability, state transitions, and async ordering.
  • Data and security: authorization, input handling, secret exposure, unsafe writes, and concurrency.
  • Contracts: API compatibility, error semantics, lifecycle cleanup, and dependency direction.
  • Maintainability: misleading names or comments, unnecessary indirection, duplication, and related dead code.
  • Performance: concrete expensive work or regressions on the affected path; do not request memoization without a reason.
  • Validation: missing evidence for changed behavior under the chosen test mode, not missing tests merely because code changed.

For each actionable finding, include the affected file and location, failure conditions, consequence, supporting evidence, and a suggested correction. Separate confirmed defects from unresolved questions. Pre-existing issues outside scope and subjective preferences should not become blockers.

Independent review

Use an independent reviewer when --peer is requested, or when complex or high-risk changes need a second opinion. Do not require it for routine PRs. Give the reviewer a bounded scope, exact diff or revision, expected behavior, and relevant constraints; request findings without edits.

If available, codex-plugin-cc provides /codex:review for branch review and /codex:rescue for targeted analysis. Use capabilities actually installed in the current environment; do not assume a plugin or specific tool is present. If an explicitly requested reviewer is unavailable, disclose the limitation and complete the useful local review.

Run local review alongside an independent review only when they can proceed independently. Integrate the returned findings before declaring the requested peer review complete. Verify each finding rather than accepting it because of reviewer agreement or confidence.

Resolve findings

  • In fix-authorized work, apply confirmed corrections within scope, including logic changes whose intended behavior is established.
  • Follow the global authorization boundaries for API or schema changes, destructive actions, and material scope changes. Ask only when a required decision remains unresolved; do not ask separately for every logic edit.
  • For review-only work, report defects and proposed corrections without making edits.
  • Reject incorrect feedback with concrete evidence. Avoid adding speculative abstractions or behavior to satisfy a reviewer.

After corrections, inspect the final diff and validate according to the global test mode and required project checks. Do not repeat checks without a relevant change or unresolved concern.

Completion

Lead with actionable remaining findings, ordered by impact. Otherwise state that no actionable defects were found within the reviewed scope. Summarize fixes made, checks actually performed, and material limitations; absence of findings is not proof that no bugs exist. Keep the report proportional to the change and do not imply a peer review ran when it did not.

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.