1. Browse public Skills
Use Marketplace or Registry to find a Skill by category, runtime, verification state, permissions, or publisher.
5-minute setup
Move from public discovery to a controlled runtime call without guessing which pieces are live and which are still preview.
Use Marketplace or Registry to find a Skill by category, runtime, verification state, permissions, or publisher.
Open the manifest before installation. Check schema, runtime, permissions, version, examples, and review state.
Sign in, create a project, and issue a scoped key for approved runtime calls.
Invoke the Skill through the runtime endpoint with Authorization: Bearer PROJECT_KEY and a typed input payload.
Add SkillHub MCP to your Agent workbench so approved Skills appear as callable tools.
Check runtime logs, policy blocks, rate limits, and preview billing signals after test calls.
Choose one low-risk workflow first, such as SEO page review, listing QA, CSV cleanup, or support ticket summary. Avoid customer secrets and production writes during the first test.
A first REST test should prove that authentication, input schema, response shape, policy checks, and logs all work before you automate anything.
A first MCP test should show that the Agent workbench can discover approved tools, read tool descriptions, and call only the Skills allowed by project policy.
Most early failures come from missing Project Keys, wrong environment variables, schema mismatch, policy blocks, or calling a preview-only Skill as if it were production-ready.