SuperAPI AI Directory

adversary-run

Security & auditing

View repository

Updated 10-23 23:07

About this skill

Your objective is to implement $adversary run [--target <path|stage|plan>] [--security] [--complexity].

Run a read-only challenge pass that looks for incorrect assumptions, weak evidence, hidden dependencies, security risk, complexity debt, and lifecycle gate bypass. Write the report under .cocoplus/adversary/reports/.

Anti-Rationalization

Shortcut / Temptation Why It Fails
Treat the skill as complete because the file exists Skill contracts must describe observable behavior and verification, not just command names.
Skip artifact and safety checks for a small command Small commands still mutate state or guide execution; preserve the same gates.

Exit Criteria

  • Command behavior matches the owning feature contract.
  • Required reads, writes, and user-visible outputs are described.
  • Safety, governance, and artifact constraints are preserved.
  • Missing state produces a clear, non-destructive result.