adversarial-reviewer

Adversarial code review that assumes bugs exist and hunts for them. Use when asked to review code, find bugs, audit for correctness, stress-test a PR, or when someone says "tear this apart" or "what's wrong with this". Give no benefit of the doubt — every line is guilty until proven innocent.

Quality

86%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

Passed

No findings from the security scan

Quality

Content

78%Weight 40%Scale 1-5

Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.

A well-structured instruction skill that is lean, actionable, and clearly sequenced, with a concrete output template and ordered checklist. It sits just below the top tier on conciseness, actionability, and workflow clarity because its validation/checkpointing is implicit rather than templated.

Suggestions

Add an explicit validate-then-act checkpoint in the Process section (e.g. 'Before writing up, re-confirm each finding with a concrete trigger') to lift workflow_clarity toward 5.

Tighten the Mindset prose ('Your job is to find bugs, not to be helpful' and the 'no sandwich' line) into bullets to push conciseness to the lean 5 anchor.

Optionally include one short filled-in BUG example in the Output Format so the template is unambiguously copy-paste ready.

Dimension	Reasoning	Score
Conciseness	Bullet-driven and free of concept padding, with only minor stylistic atmosphere ('There is no sandwich. There is only bugs.') beyond the lean anchor; efficient but not quite token-perfect at level 5.	4 / 5
Actionability	Concrete checklist items (e.g. 'Off-by-one in loops, slices, ranges', 'bare catch / catch(e)') and a copy-paste output template give mostly executable guidance for an instruction skill, with only minor gaps versus the fully-ready level 5.	4 / 5
Workflow Clarity	An ordered 'Work through these categories in order' checklist plus a 5-step Process with explicit feedback ('If you found nothing, say No bugs found and stop') gives a clear sequence, though validation checkpoints are softer than the validate->fix->retry level-5 anchor.	4 / 5
Progressive Disclosure	A single-purpose, tightly sectioned skill with no external bundle and clear headers (Mindset, Review Checklist, Output Format, Severity Guide, Process) matches the well-organized-simple-skill case that earns a 5.	5 / 5
	Total	17 / 20 Passed

Description

95%Weight 40%Scale 1-5

Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.

A crisp, third-person description that cleanly answers both what and when with concrete, natural-language triggers and a distinct adversarial niche. Slightly shy of perfect specificity only because the action list could be marginally more comprehensive.

Dimension	Reasoning	Score
Specificity	Names the domain and several concrete actions ('hunts for them', 'find bugs, audit for correctness, stress-test a PR') with only minor coverage gaps, matching the 'lists several specific actions' anchor rather than the fully comprehensive level 5.	4 / 5
Completeness	Explicitly states both what it does ('Adversarial code review that assumes bugs exist and hunts for them') and when to use it with concrete trigger phrases ('Use when asked to review code, find bugs...'), matching the top anchor.	5 / 5
Trigger Term Quality	Covers natural user phrases and synonyms including quoted 'tear this apart' and 'what's wrong with this' alongside 'review code, find bugs, audit for correctness, stress-test a PR', matching the comprehensive-synonyms anchor.	5 / 5
Distinctiveness Conflict Risk	Carves a clear adversarial-review niche with distinctive triggers ('tear this apart', 'guilty until proven innocent') that minimize overlap with a general code-review skill, matching the minimal-conflict anchor.	5 / 5
	Total	19 / 20 Passed

Validation

100%

Warnings & errors only

Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.

Validation — 16 / 16 Passed

Validation for skill structure

No warnings or errors.

Repository: emdash-cms/emdash
Path: skills/adversarial-reviewer/SKILL.md
Commit: 1717d31

Reviewed: 2 days ago

Table of Contents

Discovery Implementation Validation

Is this your skill?

If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.