Adversarial review

Skill IrfanSadiqRahat/constellation/methodology/adversarial-review

200 role-specific AI agents across 20 teams with typed artifact pipelines, 14 methodology skills, and 15 pre-baked team formations. The virtual engineering org for Claude Code, Cursor, Codex CLI.

Install

npx -y skills add IrfanSadiqRahat/constellation --skill adversarial-review

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

2 things to look at

no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.
1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Three competing reviewer roles produce independent verdicts in parallel; coordinator synthesizes.

SKILL.md

2.2 KB, as published. Nobody here has run it

adversarial-review

When activated

High-stakes work needs hostile review: security-sensitive code, irreversible changes, public APIs, payment flows, AI prompts that move money.

Single-reviewer review is biased. Adversarial review uses three roles in parallel with different lenses.

The three lenses

reviewer	optimizes for	hates
`code-review-and-quality`	clarity, maintainability, idioms	clever code, deep nesting, weak names
`security-and-hardening`	adversarial inputs, trust boundaries	implicit trust, missing validation, secret leak
`performance-optimization`	latency, memory, cost	n+1, allocation in loops, blocking on critical path

The three run independently. They don't see each other's verdicts.

Process

1. Brief each reviewer with the artifact + the role-specific lens
2. Each produces a ReviewReport with severity-tagged findings
3. Coordinator role (`code-review-and-quality`) aggregates:
   - Critical from ANY → critical
   - High from 2+ → critical
   - Each reviewer's lens preserved in the report
4. Producer addresses critical + high
5. One more round of all three until zero critical

Verdict format

artifact: <id>
reviewers:
  - lens: quality
    findings: [<severity, location, fix>]
  - lens: security
    findings: [...]
  - lens: performance
    findings: [...]
synthesis:
  unanimous_critical: [...]
  contested_high: [...]   # at least one says high, another says low
  blockers: [...]
recommendation: merge | revise | redesign

When NOT to use

Low-stakes refactors
Bug fixes with clear scope
Internal-only experiments

Three reviewers cost 3× a single reviewer. Use accordingly.

Why this skill is opinionated

Three independent lenses — catches biases of a single reviewer
Coordinator synthesises — producer gets one verdict, not three opinions to interpret
Contested findings surfaced — high-signal cases for human judgement