Uncork

Both tools query the same product index, so their surface is inherently close, but the descriptions draw a clear line: find_fix takes a natural-language problem statement and returns LLM-ranked picks, while search_products takes 3-16 keywords and returns unranked raw matches. An agent can reasonably choose based on whether it has a problem description or just keywords.

Both names are snake_case verb_noun pairs (find_fix, search_products) with an obvious, predictable pattern. The verbs differ (find vs search) but both accurately reflect their respective behaviors rather than breaking convention.

Two tools is on the thin side for a discovery service; the entire surface is two variations of 'search the index'. Each earns its place, but the set feels minimal rather than well-scoped.

The core discovery workflow is covered in both ranked and raw modes, but there is no way to fetch full details for a specific product hit, browse by category, or otherwise follow up on a result. Agents can work around this by reading the returned links, but the surface stops at search.