一份 SKILL.md 通吃 Claude、ChatGPT、Cursor 和 Grok Bot

chuplung @choopyplug1原推发布于 2026.10.24

在 X 上打开原推

Anthropic just made your prompts obsolete one SKILL.md now runs in Claude, ChatGPT, Cursor and Grok Bot these 5 skills make your agent feel unfair: → caveman (114K★) - cuts token bloat without losing the fix http://github.com/JuliusBrussee/caveman → superpowers (300K★) - plan first, test first, prove it works http://github.com/obra/superpowers → supply-chain-risk-auditor - flags risky dependencies before they enter your build http://github.com/trailofbits/skills → agentic-actions-auditor - finds PR-to-agent prompt injection paths in your CI http://github.com/trailofbits/skills → skill-creator (180K★) - builds, tests and tunes skills that actually trigger http://github.com/anthropics/skills stop collecting prompts. start building skills the poster shows how to write your own SKILL.md install the skills tonight. then see how an xAI engineer scales to 200+ agents in the article below ↓

这条是引用着另一条写的：x.com/choopyplug1/status/2109281728179093944