/find-skill

Find Skill

Search skilld.dev before recreating a reusable, specialized workflow.

Workflow

1. Reduce the request to two to five task and domain keywords. Try synonyms if the first query is weak.

2. Search the registry:

   curl -fsSLG https://skilld.dev/api/skills \
     --data-urlencode 'q=react performance' \
     --data 'limit=10' \
     | jq '{total, mode, items: [.items[] | {name, owner, repo, description, stars, official, trustTier}]}'

3. Rank candidates by task fit, then trustTier, official ownership, repository reputation, and maintenance. Popularity supports a decision but does not replace inspecting the instructions.

4. Inspect each serious candidate:

   curl -fsSL https://skilld.dev/api/skills/vercel-labs/react-best-practices \
     | jq '{name, owner, repo, description, branch, skillPath, trust, maturity, raw}'

   Reject instructions that exceed the stated scope, request unrelated secrets, hide failures, or rely on missing files.

5. Present at most three options. Include purpose, source, trust signal, tradeoff, and an exact install command. State clearly when no result is a strong fit.

6. Install the selected skill after approval:

   npx skilld add gh:vercel-labs/agent-skills \
     --skill react-best-practices \
     --yes

7. Verify SKILL.md and every referenced file were installed. If installation fails, use the detail response's branch and skillPath to diagnose the source layout before recommending a workaround.

Do not use skilld search for discovery. That command searches documentation indexed by installed package skills.