Check reference / Triggering & Discoverability
TRIGGER-01 Description triggers reliably
LLM-assisted — runs with an Anthropic key (BYOK on the CLI, managed on Pro hosted scans).
Why it matters
The description is the ONLY text the model reads when deciding whether to invoke your skill — if it doesn't match what users actually type, the skill silently never fires. This is the #1 real-world skill failure: our ecosystem scan found ~73% of public skills have descriptions unlikely to trigger reliably.
How to fix it
Lead with the concrete use case, then add 2–3 natural-language phrases a user would actually say. Keep the scope matched to what the body really does — over-broad descriptions misfire on unrelated prompts, which is just as fatal. An LLM judges the description against the body (pass at 80/100).
Example
✗ flagged
Helps with documents.✓ passes
Converts Word and PDF files to clean Markdown. Use when the user says "convert this doc", "extract the text from this PDF", or attaches a .docx/.pdf they want as text.