This skill is part of the claude-music plugin. Only invoke when the user explicitly uses the slash command.
Vibe
Auto-detect the right music based on what's happening in the current session. No user input needed — the DJ reads the room.
Instructions
You have the full conversation context. Pick the best music and play it in ONE step. Do NOT spawn an agent or delegate.
Step 1: Read the room and pick music
Read the genre metadata from sources.yml:
python3 -c "
import sys
try:
import yaml
with open('${CLAUDE_PLUGIN_ROOT}/config/sources.yml') as f:
data = yaml.safe_load(f)
except ImportError:
sys.exit(1)
for genre, info in data.items():
print(f'{genre}:')
for s in info.get('stations', []):
tags = ', '.join(s.get('tags', []))
print(f' - {s[\"name\"]}: {s.get(\"description\", \"\")}')
print(f' tags: {tags}')
print()
"
Using the tags and descriptions, match the session context to the best genre. Consider:
- What the user has been working on (e.g. debugging, building a feature, reviewing code)
- The general mood/pace of the session
- Any explicit mood signals from the user
When in doubt, lean towards relaxing genres (lofi, ambient, classical). The user prefers a chill atmosphere. Only pick high-energy genres if the session strongly suggests it.
Step 2: Play it
Pick a specific station name from the matched genre, then:
"${CLAUDE_PLUGIN_ROOT}/scripts/music-controller.sh" play "<station name>"
Step 3: Respond
Return a message in EXACTLY this format:
♪ Now playing {genre} — {station name}. {one-liner reason tied to the session} ♪
- If
takeoveristruein the JSON output: prefix with Switched over from another session — - If error about no audio player: run the
install_command(ifhas_sudois true) ornosudo_hint(if false) from the JSON, then retry play automatically.
Keep it to one line. Do not explain your reasoning.