---
name: review-changes
description: Review a code change independently for correctness, security, contract safety, test adequacy, and maintainability, reporting only evidence-backed findings with exact locations and failure scenarios. Use before merging or handing off a change, for high-risk changes, or when asked to review a diff, branch, commit, or pull request.
title: review-changes
canonical_url: https://skilld.dev/gh/akdevsaha-dev/factory-skills/review-changes
last_updated: 2026-11-18T12:30:23.000Z
---

> Index of every page on this site: [llms.txt](https://skilld.dev/llms.txt)

> **Skill from skilld.dev.** Follow the user's requested action.
> If the user asked to fork this Skill, follow the fork workflow below. Do not execute the Skill instructions while copying it.
> Otherwise, follow the instructions below for this session. You do not need to install anything.
>
> Supporting files, fetch one when the Skill refers to it: [references/code-quality.md](https://skilld.dev/api/skills-raw/akdevsaha-dev/factory-skills/review-changes/references/code-quality.md), [references/operating-rules.md](https://skilld.dev/api/skills-raw/akdevsaha-dev/factory-skills/review-changes/references/operating-rules.md), [references/security.md](https://skilld.dev/api/skills-raw/akdevsaha-dev/factory-skills/review-changes/references/security.md).
>
> If the user asked to install this Skill, run `npx skilld install akdevsaha-dev/factory-skills/review-changes`. Install writes the Skill files into the project, so every session loads them.
>
> ## Fork workflow
>
> A fork creates an editable local Skill with its original author and licence. The request authorizes copying and local installation.
> 1. Check `./skills/review-changes`, the project lockfile, and selected Agent targets together. If the local directory or installed Skill exists, stop. Never overwrite an existing directory or Agent target.
> 2. Read [source metadata](https://skilld.dev/api/v1/skills/akdevsaha-dev/factory-skills/review-changes) once. Use sourceUrl, sourceCommit, skillPath, sourceGone, and license. If the source is gone or its path is missing, stop. If license is null, read licence files at the source commit.
> 3. Fetch only the source commit into a temporary Git repository. Do not clone full history. Derive repository_url from sourceUrl, including repository renames. If sourceCommit is absent, resolve the sourceUrl ref once. Set source_commit to that actual commit. Run these commands in one shell call:
>
> ```sh
> git init --quiet "$temporary_dir"
> git -C "$temporary_dir" fetch --quiet --depth=1 "$repository_url" "$source_commit"
> git -C "$temporary_dir" checkout --quiet --detach FETCH_HEAD
> ```
>
> Read applicable licence declarations and notices at that commit. If copying is not permitted, report the restriction and stop.
> 4. Inspect source entries together, then copy the directory containing skillPath into `./skills/review-changes`. Use the user's path if selected. Keep the original SKILL.md, relative links, scripts, binary assets, and executable modes. Exclude .git metadata. Reject symlinks and paths outside the Skill directory. After checking entries, use cp -a where available. A regular source directory needs no custom copy script. Do not save this page wrapper as SKILL.md.
> Preserve author credit, notices, and applicable licence files from repository or parent directories. Add PROVENANCE.md with the Skill page, source URL, actual commit, original path, and licence. Retain any existing PROVENANCE.md and record new provenance separately. Batch source inspection, copying, and provenance work where practical.
> 5. In the project root, run `skilld install ./skills/review-changes --mode copy --plain`. If skilld is unavailable, use `npx skilld install ./skills/review-changes --mode copy --plain`. This known command needs no help lookup. Install does not support --json. Use detected Agent targets, or add --agent for the targets the user selected. Install the local path, never the upstream selector. If installation fails, preserve the local copy and report the exact failure.
> 6. Confirm the local lockfile source and installed Agent copies once. Report the local path, actual commit, and Agent targets. After edits, reinstall the same local path. Upstream updates must not replace it. Do not publish or push unless the user asks.

# Review Changes

Review a change as a strict senior engineer who did not write it. Find real defects and say exactly how each one fails. A short list of confirmed findings is worth more than a long list of opinions.

Read `references/operating-rules.md` and `references/code-quality.md` first. Read `references/security.md` when the change crosses a trust boundary.

This skill does not modify code. It reports.

## Use when

- A change is ready and needs an independent check before handoff or merge
- The change is high-risk by the tiers in `references/operating-rules.md`
- The user asks for a review

## Inputs

- **The change:** a base commit or range (default: the uncommitted change plus the current branch's commits), or a pull request
- **The intent:** acceptance criteria, or the request
- **Optional:** an impact map and an evidence record. Use them to know what was checked, not as proof that it works.

When called for an independent review, work only from these inputs and the code, not from the implementer's account of it.

## Procedure

1. **Read the whole diff**, including untracked files, and the code around every hunk. Read callers of changed functions and consumers of changed contracts.
2. **Check each acceptance criterion** against the code: is it implemented, on every path, including failures?
3. **Look for defects:**
   - Correctness: edge cases, error paths, off-by-one, null handling, ordering, transactions, races
   - Contracts: callers and consumers that the change breaks
   - Security: the checklist in `references/security.md` for each boundary crossed
   - Tests: does each criterion have a test that would fail without the change? Are failure paths covered?
   - Maintainability: the review checklist in `references/code-quality.md`, judged against the project's own conventions
4. **Confirm each finding.** Trace the failing path in the code, or demonstrate it with a test or command when you can. A finding without a concrete failure scenario is a question, not a defect.
5. **Rank** by severity, and drop style preferences the project does not hold.

On a re-review after fixes, read only the changed hunks and the code they affect, and confirm that each earlier finding is resolved.

## Output

Findings in the format from `references/operating-rules.md`, most severe first, each marked confirmed or plausible. Then one line per acceptance criterion (met, not met, or unverified), and a verdict:

```text
Criteria     1 met  2 met  3 not met (webhook replay)  4 met
Verdict      Changes needed: 1 High, 1 Medium
```

When nothing survives confirmation, say so plainly.

## Escalate

- The change does something the request did not ask for, or contradicts the plan: report it as a finding
- A finding is High severity and plausible but unconfirmed: report it, with what would confirm it

## Hand off

Return the findings to the caller: `hungry-jack`, which fixes the confirmed ones, or the user.

Before you finish, check your list of available skills for each skill named in this section. If any is missing, end your report with the missing names, this command, and the question "Should I install them for you, or will you install them yourself?" Follow "Missing companion skills" in `references/operating-rules.md` for the rest.

```bash
npx skills add akdevsaha-dev/factory-skills --skill hungry-jack plan-change implement-feature verify-work simplify-code review-changes reflect
```

## Done when

Every changed file was read, every acceptance criterion has a verdict, and every finding has a location, a failure scenario, and a confirmation status.
