receiving-code-review

name	receiving-code-review
description	Use when review comments, pull-request findings, audit observations, or change requests arrive. Verifies each finding before accepting, rejecting, fixing, or closing it.
activation	code review response, reviewer feedback, audit finding triage, or requested changes.
required_inputs	Review finding, cited range, approved requirements, and a safe verification surface.
required_outputs	Evidence-backed finding disposition, repair receipt when needed, and re-review request.
authority_dependencies	00_start_here/SOURCE_OF_TRUTH.md; skills/test-driven-development/SKILL.md; skills/verification-before-completion/SKILL.md.
evidence_requirements	Executed probe or cited contract evidence for every accepted, rejected, or deferred finding.

Receiving Code Review

Respect a reviewer without treating every comment as automatically correct. Verify the finding against the current change and the approved requirement.

Finding Decision Table

Outcome	Required response
Valid defect	Add a focused regression or evidence probe, fix it, and request re-review.
Valid but scope change	Return to plan/owner for an explicit decision; do not smuggle it into the patch.
Misunderstanding	Provide the exact source, test, or contract that resolves it and keep the review open for confirmation.
Disagreement	State the trade-off, evidence, owner, and release impact. Do not dismiss by authority.
Blocked	State the unavailable evidence and the one action needed to resolve it.

Process

1. Restate the finding as a falsifiable claim.
2. Inspect the cited range and run the smallest probe that could prove it wrong.
3. Classify using the table. Record evidence, not confidence alone.
4. For a valid defect, invoke test-driven-development: Red, Green, Refactor.
5. Re-run the original reviewer evidence and request re-review of the fixed range. Do not mark it closed from the implementer’s own result.

Worked Example

Finding: “The installer can overwrite a user README.”

Probe: create a temporary project with an unowned README.md, run install, and assert the command exits before writes. If it overwrites, the finding is valid: add the regression first, fix collision detection, and give the reviewer the exact command and result. If the probe already fails safely, cite that result and ask the reviewer whether a different path was intended.