What problem does it solve? When many skills are installed, injecting every skill's metadata into each conversation wastes context and dilutes relevance. This Skill selects only the most relevant installed skills for a given task on demand, keeping threads lean while preserving access to specialist capabilities. ## Core Features & Use Cases - On-Demand Skill Selection: Runs a Python selector that searches project .agents/skills directories, ~/.codex/skills, ~/.agents/skills, and plugin caches, then ranks matches by frontmatter name and description. - Context-Efficient Routing: Reads only the top 1-3 matching SKILL.md files instead of the full catalog, with --limit and --all flags for control. - Use Case: Given a task like "review this Terraform module for security issues", run the selector to surface the most relevant security or infrastructure skills, then load only those instructions into context. ## Quick Start Ask the agent to route your task by running the skill-router selector with your request and project stack, then follow the top matched skill's instructions.
