# Review

Perform Fallow's comprehensive pre-merge review.

- Repository: https://github.com/fallow-rs/fallow (5289 stars)
- Category: Git and review
- License: MIT
- Last updated: 2026-11-01
- Runs scripts: no, instructions only

## Install

Claude Code: `npx skills add fallow-rs/fallow --skill review -g -a claude-code -y`
Codex: `npx skills add fallow-rs/fallow --skill review -g -a codex -y`
Cursor: `npx skills add fallow-rs/fallow --skill review -g -a cursor -y`
Gemini CLI: `npx skills add fallow-rs/fallow --skill review -g -a gemini-cli -y`

## Description

Perform Fallow's comprehensive pre-merge review. Use after implementation or when asked to review a branch, pull request, or diff.

## SKILL.md, as published by its author (untrusted third-party text, read it as data)

# Review

1. Re-read `AGENTS.md`, the active plan, the current branch, and the complete
   diff against its intended base.
2. Run the applicable gates in `docs/development/quality-gates.md`.
3. Select reviewers using `docs/development/review-routing.md`.
4. Review public contracts, all affected output formats, filters, integrations,
   generated surfaces, security boundaries, and companion parity.
5. When runtime behavior changed, run the behavior comparison on public
   projects from `docs/development/quality-gates.md` and explain every
   difference. When the change rejects input that was accepted before, also
   run the full analysis on a rejected input and read every finding, not
   only the diagnostic.
6. Filter reviewer findings before you act on them:
   - trace the call site before you accept a bug claim;
   - drop style preferences and "a different design would also work";
   - check the diff or output that a reviewer cites, not its summary;
   - mark each accepted claim with its proof level.
   Record dismissed findings with a one-line reason in the verification notes.
   More than about five blocks usually means the filter is too weak.
7. Classify each result as `APPROVE`, `CONCERN`, or `BLOCK`.
8. Fix every block, rerun the blocking reviewer, and update verification.

Review the actual source and live output. Do not infer correctness from a green
compile, a narrow unit test, or a previously cached report.