Kappaemme-git/codex-complexity-optimizer: does the agent reach for it? 2026-10-11
Twelve Codex skills were installed together and each agent was sent three requests meant for this skill (complexity-optimizer), in a throwaway sandbox. A run stopped as soon as a skill was opened.
Then the same requests with 100 and 500 other real skills installed beside the twelve. Codex ran gpt-6-astra, Claude Code ran Claude Opus 5.5.
All levels · The requests · How a run is scored
The least reliable to trigger: it was also the one miss at 100 on both agents.
| Request | Codex, 0 others | Codex, 100 others | Codex, 500 others | Claude Code, 0 others | Claude Code, 100 others | Claude Code, 500 others |
|---|---|---|---|---|---|---|
| asked plainly Analyze this codebase for algorithmic complexity and performance hotspots, then implement safe optimizations that preserve behavior, APIs, outputs, and tests. Pay special attention to nested loops, repeated scans, unnecessary recomputation, and inefficient database access. | opened it 6s | no skill opened finished | opened it 145s | opened it 2s | opened it 2s | opened it 3s |
| described without the skill's words Review the repository to find the parts that waste the most time or resources, and improve them without changing what users or callers observe. Keep the existing tests and interfaces working, and make the changes maintainable. | opened it 6s | opened it 8s | no skill opened finished | opened it 3s | no skill opened finished | opened it 3s |
| in the other language 请分析这个代码库中的算法复杂度和性能瓶颈，并在不改变行为、API、输出结果和测试的前提下实施安全优化。重点检查嵌套循环、重复扫描、不必要的重复计算以及低效的数据库访问。 | opened it 22s | opened it 8s | no skill opened finished | opened it 2s | opened it 3s | opened it 2s |