skill-finderlisted

# Skill Finder Find and verify the best agent-usable skill, MCP server, connector, workflow, CLI, package, documentation route, hosted service, model, dataset, or composed capability stack for a real task. Skill Finder uses one adaptive evidence engine with two interoperable lanes: - **Public/local lane:** works without accounts using local source, public registries, public APIs, official documentation, and host-provided public tools. - **Connected lane:** adds private repositories, authenticated services, hosted research, enterprise code intelligence, and paid execution only when those routes are available, authorized, and relevant. Use a **task-specific evidence floor**. Missing optional routes do not block a run. They lower evidence coverage, trigger a different recovery route, or remain as Unresolved Research Lines. Stop only when the available routes cannot verify the claims needed for a responsible recommendation. ## Workflow 1. Infer the real skill need, target host, data sensitivity, cost limit, and output mode. Ask one clarifying question only when its answer would change the search lane, source family, or winner. 2. Choose **Quick Evaluation** for a straightforward, low-risk lookup. Choose **Deep Evaluation** for multiple candidates, repository or API tracing, costly or complex setup, no-good-skill work, or an explicit research request. 3. Build a capability snapshot. Verify the local essentials `python3`, `git`, and `rg`. Record other r