SkillHub Scenario Matcher

Problem

A natural-language request often cannot be mapped directly to a searchable Skill name. A user may ask for “automated weekly reports to my manager,” while the relevant Skill may be named weekly report, work summary, or tagged under office-efficiency. Searching once with the original phrase can miss useful results or mix weakly related items into the output.

How It Works

The skill turns Skill discovery into a structured workflow:
- Scenario extraction: pull out task intent, likely category, Chinese and English keywords, and expand them with synonyms or broader terms.
- Repeated search: query multiple candidate terms against the public API, merge results, and deduplicate by slug to avoid duplicate candidates.
- Intent ranking: use name and description to judge relevance, filter weak matches, and select the three to five best fits.
- Recommendation output: present each candidate with a matching reason and ask the user to choose; the list stays limited to candidates without installation actions.
- Post-selection install: run installation only after the user selects a specific item.

It fits cases where a vague request needs to be narrowed into installable Skills. The boundary is that recall still depends on keyword coverage and API results; for niche scenarios, synonyms, broader terms, or a relaxed category filter may be needed.