01The problem
AI agent skills fail silently: a vague description means the skill never fires — or worse, it fires on prompts it was never meant for. Before this tool there was no proper way to test a SKILL.md before shipping it.
AI Tooling · Testing · 2026 · LIVE
trigger-doctor
Testing for AI agent skills: it tells you why a skill isn’t triggering (or fires too often), suggests a fixed description, and saves a test suite. One SKILL.md format, 79 agents supported.
The story
The problem, how I built it, and what came out of it.
02How I built it
You run it with npx: it checks a SKILL.md against a set of prompts, shows which ones trigger and which just miss, suggests a description fix, and saves a test suite so future edits can’t quietly break the skill.
03The outcome
Listed on the skills.sh directory, passing 3/3 published audits, with support for 79 agents under one SKILL.md format.
Want something like this for your business?
The fastest way is WhatsApp — I usually reply within hours.