---
name: red-team-adversarial
description: Adversarial security and resilience analysis — auto-triggered during /review and /test based on task classification (hotfix=Lite, feature=Full, architecture-change=Full+Beast Mode). Complements OWASP security scan.
phases: ["review", "test"]
trigger: auto (classification-based)
trigger_priority: contextual
load_policy: on-match
cost_type: mixed
cost_risk: high
runtime_anchor: [".agent/workflows/review.md#Red Team Scan", ".agent/workflows/test.md#Step 4"]
---

# Red Team / Adversarial Testing

Auto-triggered adversarial analysis skill. Full instructions: `.agents/skills/red-team-adversarial/SKILL.md`.

## Quick Reference

- **Trigger**: Auto — reads classification from Work Log. Skips tiny-fix/quick-win.
- **Runtime anchor**: `/review` and `/test` decide whether to load the heavy adversarial skill body
- **Token note**: Output cost is usually larger than read cost because findings can expand across review, test, and ship
