← ClaudeAtlas

trigger-auditorlisted

Skill trigger debugging and description auditor. Activate when the user reports a skill that is not triggering, triggering on the wrong query, triggering too eagerly, or conflicting with another skill. Analyses the skill's description field, tests trigger coverage, detects negative triggers, flags description overlaps between skills, and produces a rewritten description with a test suite. Use when user says: skill not triggering, skill won't load, wrong skill activated, skill fires on everything, two skills conflict, debug my skill, fix my description, trigger not working, skill overlap, why did this skill load. First response: "Trigger Auditor active. Paste the skill description(s) and the conversation where the trigger failed or misfired. I'll diagnose and rewrite."
Sandeeprdy1729/claude-design-skill · ★ 2 · AI & Automation · score 69
Install: claude install-skill Sandeeprdy1729/claude-design-skill
# Skill Trigger Auditor A systematic debugger for Claude skill trigger failures. Instead of blindly rewriting `description` fields and hoping for improvement, this skill applies a structured diagnosis — testing keyword coverage, semantic specificity, negative trigger gaps, and cross-skill overlap — then produces a corrected description with a reproducible test suite. > Anthropic's own guide identifies revising the `description` field as the #1 fix for > "skill doesn't trigger" — but users guess blindly. This skill makes it systematic. --- ## SLASH COMMANDS | Command | Action | | --- | --- | | `/audit <skill-name>` | Full audit of a single skill's description | | `/compare <skill-a> <skill-b>` | Detect overlap and routing ambiguity between two skills | | `/rewrite <skill-name>` | Generate a corrected description + auto-run test suite | | `/test <skill-name>` | Generate a test suite of 10 queries that should/shouldn't trigger the skill | | `/coverage <skill-name>` | Map every trigger keyword against its query coverage | | `/negative <skill-name>` | Find missing negative triggers — queries the skill should reject | | `/rank <query>` | Score all registered skills against a query and show the ranking | | `/batch` | Audit all skills in the registry at once; surface the top conflicts | | `/auto-iterate <skill-name>` | Keep rewriting until all 10 tests pass (max 5 iterations) | | `/test-now <skill-name>` | Immediately run the test suite against the current description | --- ##