← All agent skills

Skills-compatible agents / Code review

Evaluating Code Review Feedback

Checks review suggestions against real code, compatibility constraints and user decisions before implementing them.

315,272 repository stars · checked 2026-11-10. Count applies to obra/superpowers, not this individual skill.

Guide by IDAKIM Digital. Source review: . Original explanation based on the linked publisher sources.

What it does

The skill focuses on the receiving side of review: understand, verify, evaluate and respond with technical substance. It discourages automatic agreement and bundled edits made before the reviewer’s meaning is clear. External feedback is checked against actual usage, supported platforms and the reasons existing behavior exists. Suggestions for unused infrastructure face a YAGNI check rather than automatic expansion. Valid corrections are implemented one at a time, with evidence that they solve the reported problem without regressions.

When to use it

  • Assess a reviewer’s recommendation that conflicts with supported OS versions.
  • Separate an actual bug from an unnecessary proposed abstraction.
  • Resolve a multi-item review containing ambiguous or interdependent requests.

Step-by-step workflow

  1. Read the entire review and restate the technical requirements before reacting or editing.
  2. Clarify ambiguous and interdependent items; check relevant code, usage and earlier decisions.
  3. Evaluate correctness for the actual platform and compatibility targets, including whether the proposed feature is used.
  4. Apply blocking fixes first, then simple and complex corrections one at a time; test each change.
  5. Respond with the verified change or a reasoned objection, correcting your own earlier assessment when evidence contradicts it.

Example request

Use receiving-code-review for these export-feature comments. Check the supported Python versions and current callers before accepting a new dependency or removing compatibility code. For each item, state the behavior at issue and evidence from the repository. Clarify anything that changes the contract. Apply valid fixes individually and rerun relevant tests. Prepare response text, but do not post it to the remote review without authorization.

Worked example

A reviewer may propose replacing a compatibility path with a newer library API. If the project supports an older runtime lacking that API, removing the path breaks real users even if the local test environment passes. The useful response cites the supported minimum and tests that boundary. If the old path is instead unused and all callers use the newer API, deletion may be correct. The evidence decides, rather than the reviewer’s confidence or the implementer’s attachment to existing code.

What you get

Verified corrections, specific technical responses and a clear account of suggestions declined or requiring clarification.

How to evaluate the result

  • Accepted suggestions address a reproduced issue and preserve required compatibility.
  • Rejected suggestions have concrete evidence rather than defensive language.
  • Tests run after corrections and the response accurately reflects what changed.

What you need

The full review, repository access, known supported environments and a runnable verification path. User decisions that constrain the change need to remain visible.

Setup and activation

The publisher documents separate installation routes for Claude Code, Codex App and CLI, Cursor and other harnesses. Use the README for your exact client; installing the whole Superpowers framework can activate additional workflows automatically. Preserve this skill’s companion files and relative references. Review and security-scan the exact selected version before installation. This directory only reviews sources; it has not installed, executed or tested this skill on those clients.

Limits and considerations

Skepticism is not license to dismiss feedback. If a claim cannot be checked, say what evidence is missing. The source asks for clarification before implementing an unclear interrelated set. Inline GitHub replies have their own thread targets and external-posting permissions; drafting a response does not authorize sending it.

Related workflows

  • Use requesting-code-review to supply a clean review brief in the first place.
  • Use systematic-debugging when a finding needs causal investigation before correction.

License and reuse

MIT, copyright 2025 Jesse Vincent. Preserve the copyright and permission notice when redistributing substantial source portions; inspect separately supplied dependencies and assets under their own terms.

Further reading

Read the original source

Publisher: obra/superpowers
Source reviewed: 2026-11-10

View SKILL.md on GitHub ↗

Source reviewed; not installed, executed or security-scanned for this directory. These example requests were written by IDAKIM Digital. Upstream content can change.