What problem does it solve? Code reviews often follow the happy path the author intended, producing false negatives, or hunt bugs randomly, producing noise. This Skill provides a systematic method to attack the assumptions behind a system before launch or refactor, turning adversarial thinking into testable hypotheses with evidence. ## Core Features & Use Cases - Persona-Based Attack Model: Six user personas (curious, malicious, power user, careless, concurrent, stale-state) each targeting a different class of system assumption. - Canonical Operations: Seven operations (repeat, reverse, reorder, skip, replay, concurrent, manipulate) that transform normal execution into pressure cases. - Evidence-Graded Findings: Confidence levels from CONFIRMED to SPECULATIVE, with false-positive screening for rate limiting, idempotency, and server-side authorization. - Use Case: Before shipping a rewards flow, apply the reverse operation to check whether undoing an action removes the granted reward, catching infinite farming bugs before production. ## Quick Start Ask the agent to run an adversarial review on a specific flow, such as "attack the assumptions behind the checkout and reward-granting flow and report confirmed findings."
