ningzimu/codex-ppt-skill: 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 (codex-ppt), 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
| Request | Codex, 0 others | Codex, 100 others | Codex, 500 others | Claude Code, 0 others | Claude Code, 100 others | Claude Code, 500 others |
|---|---|---|---|---|---|---|
| asked plainly Turn this report into a visually unified, image-based PPTX deck with concise slide text, strong visual hierarchy, and an opening, body, and conclusion: “Urban rooftop gardens can reduce building heat, support pollinators, and provide fresh herbs for residents.” | opened it 6s | opened it 10s | opened it 8s | opened it 3s | opened it 3s | no skill opened time limit |
| described without the skill's words Create a cohesive presentation made primarily from carefully selected visuals, with consistent layouts and brief captions, based on this outline: “A small coastal town restored its wetlands, reducing flood damage while improving bird habitat and recreation.” | opened it 8s | opened it 8s | opened it 20s | opened it 3s | opened it 3s | opened it 3s |
| in the other language 请将以下研究摘要制作成一套视觉统一、以图片为主的PPT演示文稿，包含清晰的开场、核心内容和结论，并使用简洁的幻灯片文字：“远程办公能够减少通勤排放，但需要改善家庭办公条件并加强团队沟通。” | opened it 8s | opened it 16s | opened it 20s | opened it 3s | opened it 4s | opened it 3s |