cyberneticlibrary

Perform hostile code reviews

adversarial-reviewerskillsetup L117,464
alirezarezvani/claude-skills
What it does

Force critical code review via adversarial personas

Best for

Catching production-breaking bugs and maintainability issues that self-review and agreeable LLMs miss

Inputs
  • · [object Object]
Outputs
  • · [object Object]
Preconditions

git repo with diffs or file to review; must read full context of changed files

Failure modes
  • · Reviewer shares same mental model as author and misses blind spots
  • · Personas produce duplicate findings that need synthesis
  • · Code genuinely bulletproof — personas must still surface fragile assumptions
Trust signals
  • · Concrete examples for Saboteur (input validation, state consistency, error paths)
  • · New Hire mindset focuses on names, logic clarity, type info
  • · Security Auditor explicitly references OWASP
  • · Mandatory finding rule ensures all three personas produce issues