Skip to content
Agent Orchestration
Skill

/receiving-code-review

Handles review feedback with technical rigor: verify suggestions, resolve unclear items, implement fixes in priority order. Invoke when review comments are received or user says "address review feedback". Routed by using-superpowers when review feedback arrives.

BOOST
From plugin
superpowers-optimized
14224 skills2 agents6 hooks
Install
$ npx -y skills add REPOZY/superpowers-optimized --skill receiving-code-review --agent claude-code

How it fires

How this skill gets triggered: by you, by Claude, or both.

  • Fires itselfAuto-invocation. Claude auto-loads it when your prompt matches the work.Auto-invocation is when the right skill fires by itself at the right moment, driven by a FLOW.md router and a hook, instead of you invoking it by name. It is the difference between a skill being installed and a skill actually getting used.Read the full definition →
  • You can call itInvoke it directly when you want it.
  • Slash command/receiving-code-review

Context preview

The summary Claude sees to decide when to auto-load this skill.

Handles review feedback with technical rigor: verify suggestions, resolve unclear items, implement fixes in priority order. Invoke when review comments are received or user says "address review feedback". Routed by using-superpowers when review feedback arrives.

SKILL.md

receiving-code-review.SKILL.md
name: receiving-code-review
description: >
  Handles review feedback with technical rigor: verify suggestions,
  resolve unclear items, implement fixes in priority order. Invoke when
  review comments are received or user says "address review feedback".
  Routed by using-superpowers when review feedback arrives.

Receiving Code Review

Treat feedback as technical input to validate, not as instructions to apply blindly.

Sequence

1. Read all feedback. 2. Clarify unclear items before implementing anything. 3. Validate each suggestion against codebase behavior and requirements. 4. Implement fixes in priority order. 5. Re-test and summarize outcomes.

Priority

1. Correctness/security regressions 2. Requirement mismatches 3. Maintainability issues 4. Minor polish

Treat Critical and High security findings as blocking until addressed or explicitly deferred by the user with documented rationale.

Forbidden Responses

Never say any of the following. They signal performative agreement, not technical engagement:

  • "You're absolutely right!"
  • "Great point!"
  • "Good catch!"
  • "Thanks for catching that!"
  • Any gratitude expression before analysis
  • Any agreement before verification

Instead: state what you verified, what you changed, and why. If the reviewer is right, the code change speaks for itself.

Pushback Rules

Push back when a suggestion:

  • Breaks existing behavior
  • Conflicts with approved architecture
  • Adds unused scope (YAGNI)
  • Lacks enough context to verify

Pushback must include **concrete technical evidence** — not opinions, not "I think", not "it should be fine."

Response Style

  • Be factual and concise.
  • State what changed, where, and why.
  • If you disagree, explain with code or test output — not with rhetoric.

Completion

Report:

  • Addressed items
  • Deferred items with reason
  • Verification commands/results
  • Remaining risks
Read more
Ships withsuperpowers-optimized

The production-grade fork of obra/superpowers — same trusted workflow, dramatically leaner, safer, and more intelligent.

Get the whole plugin

Other skills on superpowers-optimized.