Critical Code Reviewer

## Summary This skill is a high-intensity, adversarial code reviewer tailored for serious PR and code-quality workflows. It is optimized for **finding real issues**—security holes, logic errors, sloppy patterns, and maintainability problems—rather than being polite or lightweight. It aligns well with professional engineering practices where strict reviews are expected. **Key strengths** - Very strong coverage of common failure modes across Python, R, JS/TS, SQL, and front-end code. - Clear severity tiers (Blocking / Required / Suggestions) make outputs immediately actionable in PR contexts. - Explicit guidance on how to treat uncertainty and partial context (“Verify” vs “Blocking”) increases review honesty and reliability. - Focus on adversarial, real-world failure scenarios (malformed APIs, malicious input, unhandled promises, nulls) reduces risk of missing critical bugs. **Risks / limitations** - The deliberately harsh, skeptical stance may feel overly negative or intimidating for beginners or casual learning scenarios. - As with any static reviewer, it may occasionally flag “risks” that are actually acceptable by local team conventions; the skill does encourage marking such items as “Verify” but this still requires human judgment. - It is specialized for **critical review**, not for teaching fundamentals from scratch or writing new code from a blank slate. **Recommended scenarios** Use this skill when you: - Want a **brutally honest, production-grade** review before merging code. - Are explicitly asking for “critical review”, “what’s wrong with this code”, or “find issues in my PR”. - Need to stress-test security, error handling, and robustness. Avoid it when you primarily want gentle onboarding, tutorial-style explanations, or highly collaborative tone; pair it with a more pedagogical tool in those cases.