---

name: receiving-code-review description: "Use when you have received code review feedback and need to process it. [DEPRECATED] Routes to code-review --feedback" deprecated: true replacement: code-review --feedback

Receiving Code Review (Deprecated)

Invariant Principles

1. Route to Replacement - Always route to code-review --feedback
2. Pass Context Through - Forward all provided context to replacement skill
3. No Independent Execution - This skill does not execute feedback processing logic itself

Automatic Routing

When this skill is loaded, immediately invoke:

Copy/code-review --feedback

With any provided context passed through.

---

Handoff from Requesting Skill

When processing external feedback after internal review:

Context Loading

1. Check for existing review-manifest.json
2. Load internal findings for comparison
3. Cross-reference external findings against internal

Finding Reconciliation

Shared Context

Access via review-manifest.json:

• reviewed_sha - What commit was reviewed
• files - What files were in scope
• complexity - Size estimate

---

Copy/code-review --feedback

With any provided context passed through.

Migration Guide

Thread Reply Protocol

Reply Location

• ALWAYS reply in the existing thread, never as top-level comment
• Use gh pr comment --reply-to <comment-id> or MCP reply tools
• If thread ID unavailable, quote the original comment

Response Formats

FIXED - Issue addressed with code change:

CopyFixed in [commit SHA].

[Optional: brief explanation of fix approach]

ACKNOWLEDGED - Will address, not yet fixed:

CopyAcknowledged. Will address in [scope: this PR / follow-up / future iteration].

[Optional: brief plan or reason for deferral]

QUESTION - Need clarification:

CopyQuestion: [specific question]

Context: [what you understand so far]
[Optional: what you tried or considered]

DISAGREE - Technical disagreement with evidence:

CopyI see a different tradeoff here.

**Current approach:** [what code does]
**Suggested change:** [what was requested]
**My concern:** [specific technical issue with evidence]

[Optional: alternative proposal]

Happy to discuss further or defer to your judgment on [specific aspect].

Forbidden Responses

• "Done" (no context, no SHA)
• "Fixed" (no SHA, can't verify)
• "Will do" (no commitment scope)
• "Thanks!" (performative, adds no information)
• "You're right" (without explaining what you learned)

Feedback Source Trust Levels

Trust Level Actions

High Trust:

• Verify 1-2 representative findings
• Proceed with implementation if spot-check passes
• Escalate only if spot-check fails

Skeptical:

• Verify EVERY finding against codebase
• Cross-reference with internal review if exists
• Question assumptions, request evidence for vague feedback

Low Trust:

• Treat as suggestions, not requirements
• Full verification mandatory
• Escalate to partner before implementing substantial changes

Objective:

• Tool output is factual (lint errors, type errors)
• Verify tool configuration is correct
• Address systematically, don't argue with tools

MCP Tool Failures

When MCP tools fail during feedback verification, follow this fallback chain:

Failure Logging

Log every failure with:

• Tool name and operation attempted
• Error message or timeout
• Context (what verification was being performed)

Fallback Chain

1. Primary: MCP tools (pr_fetch, pr_diff, etc.)
2. Fallback 1: Direct file reading with Read tool
3. Fallback 2: Git commands via Bash (git show, git diff)
4. Fallback 3: Request manual paste from user

Hard Stop Rule

If ALL fallbacks fail for a verification:

• Report: "Cannot verify: [finding summary]"
• Do NOT implement unverifiable suggestions
• Mark finding as UNVERIFIED in response
• Escalate to user for manual verification decision

Never Implement Unverified

Why Deprecated?

The code-review skill consolidates all review functionality:

• --self: Pre-PR self-review
• --feedback: Process received feedback (this functionality)
• --give: Review someone else's code
• --audit: Comprehensive multi-pass review

See code-review/SKILL.md for full documentation.