Kappaemme-git/codex-startup-pressure-test-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 (startup-pressure-test), 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 Brutally evaluate and refine my startup idea using practical early-stage frameworks: identify whether the problem is real, analyze competitors and current customer behavior, find the first 10 customers, define an MVP, propose a two-week launch plan, assess founder-market fit, and give me a clear strong, weak, or pivot verdict. Idea: “A subscription service that helps independent restaurants reduce | opened it 7s | opened it 7s | no skill opened finished | opened it 4s | opened it 4s | opened it 3s |
| described without the skill's words I’m considering a subscription service that helps independent restaurants sell surplus ingredients to nearby home cooks. Investigate whether people truly need this, how they solve it today, who else serves them, how to reach the first 10 paying customers, what the smallest useful version should include, and what I should do over the next two weeks; finish with a clear recommendation to continue, c | opened it 8s | opened it 9s | no skill opened finished | opened it 4s | opened it 3s | opened it 3s |
| in the other language 请用实用的早期创业方法，毫不留情地评估并改进我的创业想法：判断问题是否真实存在，分析竞争者和客户目前的行为，找出前10位客户，定义最小可行产品，制定为期两周的发布计划，评估创始人与市场的匹配度，并明确告诉我是继续、改进方向，还是放弃。想法：“一个帮助独立餐馆将剩余食材卖给附近家庭厨师的订阅服务。” | opened it 19s | opened it 14s | no skill opened finished | opened it 3s | opened it 4s | opened it 4s |