agentleFS
Sign inStart free
DROOdotFOODROOdotFOO/agent-skills/skills/adversarial-reviewer/SKILL.md

adversarial-reviewer

Skill1 starsChanged 2 months agoMITView on GitHub

Devil's advocate code review with three adversarial personas that challenge assumptions and find hidden issues. TRIGGER when: user asks for adversarial review, devil's advocate feedback, wants code stress-tested, or says "tear this apart". DO NOT TRIGGER when: user wants a standard code review, quick feedback, or style-only review.

We split this skill into the rules it gives an agent

4 rules. 0 are common, 4 are distinctive, and 0 are very specific to this project.

  • 0

    Common

    Most of the 16 edited versions we read keep these.

  • 4

    Distinctive

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    When 2+ personas independently flag the same issue, promote its severity by one level:

    line 33
  2. Distinctive:

    Read the code completely before commenting

    line 43
  3. Distinctive:

    Run each persona independently -- do not let one persona's findings bias another

    line 44
  4. Distinctive:

    After review, issue exactly one verdict:

    line 50

“Common” is measured: kept by at least 6 in 10 of the 16 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Adversarial Reviewer

Review code through three hostile personas, each determined to find what others miss.

Personas

  1. Saboteur -- What breaks in production? Race conditions, edge cases, resource exhaustion, cascading failures.
  2. New Hire -- What's unmaintainable? Unclear naming, missing docs, implicit assumptions, magic numbers.
  3. Security Auditor -- What's the attack surface? OWASP-informed: injection, auth bypass, data exposure, privilege escalation.

See personas.md for detailed persona descriptions and self-review techniques.

Mandatory Findings

Each persona MUST find at least one issue. If a persona finds nothing, dig deeper -- no clean passes allowed. This forces thorough examination rather than rubber-stamping.

Severity Promotion

When 2+ personas independently flag the same issue, promote its severity by one level:

  • LOW -> MEDIUM
  • MEDIUM -> HIGH
  • HIGH -> CRITICAL

Cross-persona agreement signals systemic risk.

Review Process

  1. Read the code completely before commenting
  2. Run each persona independently -- do not let one persona's findings bias another
  3. Collect findings, check for cross-persona overlap, apply severity promotion
  4. Deliver consolidated report grouped by severity

Verdict

After review, issue exactly one verdict:

  • BLOCK -- CRITICAL or 3+ HIGH issues found. Do not merge.
  • CONCERNS -- HIGH or multiple MEDIUM issues. Merge after addressing.
  • CLEAN -- Only LOW issues. Safe to merge.

What You Get

  • A consolidated adversarial review report with findings from three independent personas (Saboteur, New Hire, Security Auditor), grouped by severity (CRITICAL/HIGH/MEDIUM/LOW).
  • Cross-persona overlap analysis with automatic severity promotion when multiple personas flag the same issue.
  • A single verdict (BLOCK, CONCERNS, or CLEAN) indicating merge readiness.

Output Format

## Adversarial Review: [file/component]

### CRITICAL
- [Saboteur] ...
- [Security Auditor] ...

### HIGH
- [New Hire] ...

### MEDIUM / LOW
- ...

### Cross-Persona Overlaps
- Issue X flagged by Saboteur + Security Auditor (promoted: HIGH -> CRITICAL)

### Verdict: BLOCK | CONCERNS | CLEAN

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

Check your setup, free