critical-review_skill

Overview

This skill delivers a strict, no-nonsense critical review of implemented work with a focus on safety, correctness, and maintainability. It frames findings into an actionable report, scores key quality dimensions, and highlights blocking issues that must be fixed before merge or release. The posture is uncompromising: mediocre code is not acceptable.

How this skill works

The skill inspects code and design against a concise checklist covering code quality, error handling, tests, architecture, and performance. It flags concrete red flags (unwraps, f64 for money, excessive clones, long functions), scores five evaluation criteria, and produces a structured review report with P0/P1/P2 priorities. Use templates for quick checks or full review reports to standardize outputs.

When to use it

• After completing an implementation to validate quality before commit
• Before a release or public deploy for a full audit
• When a pull request touches core systems or critical paths
• Whenever design or direction is unclear and you need objective critique
• Before merging changes that affect trading, risk, or live execution

Best practices

• Run the quick checklist first, then perform the detailed review if issues appear
• Score each criterion honestly; require minimum 3/5 per category
• Remove all .unwrap() and replace f64 for monetary values before approval
• Prefer small, focused functions and avoid files >500 lines
• Automate checks (clippy, tests) and attach results to the report

Example use cases

• Review a new trading strategy implementation to confirm risk controls and tests
• Audit integration code for Alpaca or Binance before enabling paper or live trading
• Evaluate refactors to ensure DDD boundaries and dependency directions are preserved
• Inspect async code paths for blocking operations and unnecessary allocations
• Validate CI failures by mapping test gaps and missing edge-case coverage