产品
trigger-doctor
AI agent skills have a hidden failure mode: runtimes like Claude Code and Cursor load only the skill's name + description line, and decide from that alone whether to open it. If it doesn't match how users phrase requests…
关于
AI agent skills have a hidden failure mode: runtimes like Claude Code and Cursor load only the skill's name + description line, and decide from that alone whether to open it. If it doesn't match how users phrase requests, the skill silently never runs. trigger-doctor tests that gate: it simulates the judgment with a labeled suite, diagnoses misses, prescribes the fix, and saves a regression suite to re-run after model updates. Zero dependencies, CI exit codes, 3/3 skills.sh security audits.
收录记录
- Product Hunt 新品
讨论这个产品