codex-complexity-optimizer

Why we included this project

For anyone running Codex on real code, this skill gives the agent a disciplined path from slow-code spotting to a concrete, reviewable report: it flags algorithmic complexity and performance hotspots, then lists each issue with file/line, current complexity, a recommended change, expected complexity afterwards, risk level, and the tests or benchmarks that would confirm the fix. The report-only default is the key difference from typical autopilot tooling: nothing is edited unless you explicitly ask it to implement the lowest-risk optimization and run the relevant tests. That keeps human sign-off in the loop for teams that want AI to propose performance work without silently rewriting production code. For an existing Codex setup, it's a straightforward add-on that makes optimization work more careful and easier to audit.