dsh-skill-matcher

You have accumulated a dozen skills and several experts in your DeepSeek Harness workspace, and the open-source catalog keeps growing. Every time a new task comes in, the first question is: which one should I call? dsh-skill-matcher answers that. It unifies locally installed skills, experts, and remote open-source entries into a single index, then performs keyword recall and relevance ranking against your natural-language requirement, returning Top-N picks with matched keywords and install instructions.

Three core capabilities: skill_matcher_scan probes multi-environment SKILL.md frontmatter and plugin.json files, merges remote catalog data, and writes a local cache (24-hour TTL, works offline); skill_matcher_match is the everyday entry point, turning a one-sentence description into a ranked recommendation; skill_matcher_list lets you browse the full inventory of local, marketplace, and open-source entries. The plugin also injects a guidance block into the system prompt so that "which skill should I use" questions are routed to the matcher automatically.

Best suited for Harness users who have built up a local skill and expert library while wanting to keep discovering new tools from the open-source ecosystem. Scanning is purely local, remote fetches are JSON catalog data only, no remote code execution, and ranking is driven solely by relevance, keeping results neutral and trustworthy.