critic_skill

Overview

This skill adopts adversarial thinking to challenge assumptions and probe reasoning without offering fixes. It focuses on identifying weaknesses, edge cases, and risks so designs and decisions get stronger before implementation. Use it when you want a focused, skeptical review rather than solutions.

How this skill works

On request it asks a series of targeted, single-question probes to expose hidden assumptions and failure modes. It follows a strict read-only rule: it only highlights problems, severity, and follow-up questions (including iterative "why" probing) and never provides corrective steps. Responses are concise, prioritized by severity, and aimed at prompting rethink or deeper validation.

When to use it

• When you need to find weaknesses in a design, spec, or architecture
• Before shipping a feature to surface edge cases and race conditions
• During threat modeling or security reviews to find attack vectors
• When a proposed optimization feels like a band-aid instead of a root fix
• To stress-test assumptions about scale, availability, or dependencies

Best practices

• Ask one focused question at a time to keep probes clear and actionable
• Use the 5 Whys: repeat "why" until a root assumption is revealed
• Flag severity (critical/important/worth considering) to prioritize issues
• Avoid proposing fixes—let the team decide tradeoffs after risks are understood
• Document the questioned assumptions so they can be revisited later

Example use cases

• Reviewing an API design to find missing input validation, concurrency gaps, or inconsistent contracts
• Evaluating a caching plan to see if it masks a deeper data or query inefficiency
• Probing a deployment strategy for single points of failure and blast radius
• Stress-testing a payment flow for race conditions, retries, and partial failures
• Challenging a performance optimization to check behavior at 10x traffic