# Flow Review

> Reviews code diffs or explicit file sets through independent correctness and QA lenses, producing a severity-ordered verdict for pre-merge audits or review gates.

- Skill: `pwdev-solucoes/flow-review` (Agent Skill, multi-file: 2 files)
- Install (CLI): `npx skillmds@latest add pwdev-solucoes/flow-review`
- Raw SKILL.md: https://api.skillmd.com/api/skills/pwdev-solucoes/flow-review/raw
- Safety review: PASS (external: skill-scanner PASS, skillspector PASS)
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools, Security, Code Review, Testing & QA
- Tags: Code Review, Diff Review, Pre Merge Audit, Qa, Test Coverage
- Author: pwdev-solucoes (https://skillmd.com/u/pwdev-solucoes)
- Updated: 2026-08-22
- Page: https://skillmd.com/skills/pwdev-solucoes/flow-review

---


# Review code and test evidence

Read [collaboration](../../references/collaboration.md), [artifacts](../../references/artifacts.md), and [safety](../../references/safety.md) before reviewing.

Review is read-only. Do not implement fixes unless the user separately requests changes.

## Establish the contract

1. Resolve the scope from an explicit file list, diff range, active Flow state, or working-tree diff.
2. Read applicable `AGENTS.md`, the relevant acceptance criteria, quality rules, and prohibitions.
3. If scope is ambiguous, inspect safely and state the bounded scope used.

## Correctness lens

- Read the complete diff, not only a summary.
- Check behavior, error paths, boundary cases, security, concurrency, data integrity, compatibility, conventions, and unnecessary complexity.
- Report only actionable findings with severity, file, tight line range, consequence, and concrete remediation.
- Use inline code comments when the runtime supports them.

## QA lens

- Map each acceptance criterion and changed behavior to existing tests.
- Run the relevant test commands when safe and non-mutating beyond normal local build artifacts.
- Identify untested failure paths and explain the smallest valuable missing test.
- Distinguish environmental failures from product failures.

Run the two lenses sequentially by default. Use separate collaborating workers only when the user explicitly requests delegation or parallel agent work, following [collaboration](../../references/collaboration.md).

## Verdict

Use the worst applicable result:

- `APPROVED`: no critical or major finding and evidence is adequate;
- `CHANGES_REQUESTED`: at least one major finding or material test gap;
- `BLOCKED`: scope, evidence, or environment prevents a responsible review.

Return findings first, ordered by severity, followed by assumptions, test evidence, and verdict. Persist a report only when the user requested it or this is an active Flow phase with artifact persistence already authorized.