Skip to content
Back to skills

Factory Review

ASecurity

Independent exact-head QA, specification review and adversarial engineering review without implementation edits.

  • 2 stars
  • 0 votes
  • 0 copies
  • 1 view
  • Added October 16, 2026
developmentcode-reviewsecurityperformance

Security analysis

A100/100

Scanned October 16, 2026

npx -y skills add Division4Studios/malt --skill factory-review --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Factory Review?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Factory Review
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/division4studios-factory-review/badge)](https://www.skillsdirectory.com/skills/division4studios-factory-review)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: factory-review
description: Independent exact-head QA, specification review and adversarial engineering review without implementation edits.
---

Inspect the pinned head/base diff, acceptance criteria and actual validation evidence. No approval may carry across a source or base change. Preserve the isolated snapshot unchanged; do not mutate code to test a hypothesis. Request a developer-owned regression or mutation experiment when needed.

QA-1 leads the specification lens: missing requirements, wrong behavior, unintended scope, user-visible edge cases, failure paths and test sensitivity. QA-2 leads the engineering lens: correctness, readable structure, architecture, security boundaries and resource/performance behavior. Both may flag any material defect independently; one passing lens cannot compensate for the other failing.

Read project standards. Distinguish enforceable violations from design judgments; avoid duplicating formatter/linter reports. Review changed behavior through public interfaces. Inspect external input validation, authorization, secrets handling, recovery, races, idempotency, cleanup ownership and unbounded operations where relevant.

Report high-confidence findings with severity (blocking / required / optional), location, triggering input, expected versus actual outcome, and reproduction or reasoning. Say which checks ran and what coverage is missing. A green suite is evidence, not proof of every requirement.

Return changes_requested for material unresolved findings. Notify through durable handoff; the supervisor returns the ticket to Development and the original developer fixes it. Never implement fixes, merge PRs or close tickets. Successful output records independent evidence for exactly the reviewed source.

Adapted from Matt Pocock code-review, Addy Osmani code-review-and-quality and Pstack prove-it-works. Their nested subagent and mutate-during-review patterns are replaced by MALT's resource-bounded immutable reviewer slots.

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.