skill-router

Claude Code orchestration skill that auto-analyzes a project on entry and recommends which installed skills to activate.

Requirements

• Claude Code (any recent version with Skill tool support)
• superpowers plugin recommended (provides 16 additional skills the router can recommend)

What it does

• Scans the repository for tech stack signals (frontend/backend/AI/SaaS/DevOps)
• Discovers installed Claude Code skills at runtime
• Recommends 3-5 relevant skills with domain-language explanations
• User confirms, then router executes skills in order (stop on first failure)

Install

git clone https://github.com/butevecom-commits/skill-router.git ~/.claude/skills/skill-router

Restart Claude Code. Done.

Optional: add a SessionStart hook to get prompted when entering a new project:

mkdir -p ~/.claude/hooks

# Create hook script
printf '#!/usr/bin/env bash\nprintf '"'"'{"additionalContext": "New project? Run /skill-router.\\n"}'"'"'\n' > ~/.claude/hooks/session-start.sh
chmod +x ~/.claude/hooks/session-start.sh

Then add to ~/.claude/settings.json:

"hooks": {
  "SessionStart": [
    {
      "matcher": "startup|clear|compact",
      "hooks": [
        {
          "type": "command",
          "command": "\"${HOME}/.claude/hooks/session-start.sh\""
        }
      ]
    }
  ]
}

Usage

/skill-router

Or: "analyze this project", "what skills should I use here?"

File structure

SKILL.md                  # Main skill entry point (6-phase workflow)
skills-index.md           # Enhancement annotations for auto-discovered skills
routing-rules.md          # Decision logic (gap-first → risk-sort → intent-override)
project-patterns.md       # Signal → project-type detection patterns
examples/
  react-project.md        # Frontend-heavy example
  saas-project.md         # Production SaaS example
  ai-agent-project.md     # AI/agent project example

Update

cd ~/.claude/skills/skill-router && git pull

Restart Claude Code.

Decisions (grilled & confirmed)

#	Topic	Result
Q1	Trigger	Hybrid: hook hints, user invokes
Q2	Index	Runtime auto-discovery + annotation file
Q3	Sort	Gap-first → risk-sort → intent-override, max 5
Q4	Confirm	Editable checkbox batch
Q5	Hook	SessionStart command hook
Q6	Mapping	Honest mapping + domain-language translation
Q7	Scheduling	Optional sequential execution
Q7b	Failures	Stop on first error

License

MIT