Default output: return only the result, blockers, and required evidence. Omit preambles, process narration, repeated context, confidence scores, and follow-up offers. Use at most five bullets unless a required artifact or schema needs more.

Thermo-Nuclear Performance Review

Use this skill for an unusually strict performance-focused review. Do not merely check for obvious N+1 queries. Actively hunt for systemic performance problems: patterns that scale poorly, hidden allocations, unnecessary serialization, and designs that become bottlenecks under load.

Above all, this skill should push the reviewer to be ambitious about performance. Do not stop at "this query could be faster." Look for fundamental design issues: O(n²) algorithms disguised as O(n), data structures that fight the access pattern, and architectures that serialize when they could stream.

Core Prompt

Start from this baseline:

Perform a deep performance audit of the current branch's changes. Evaluate algorithmic complexity, memory behavior, I/O patterns, and scalability limits. Identify every hidden cost, every unnecessary allocation, every pattern that degrades under load. Be extremely thorough and rigorous. Performance debt is invisible until production.

Non-Negotiable Performance Standards