docs: add music playback documentation

USAGE.md music section under Mumble, CHEATSHEET.md music commands,
TASKS.md sprint update for v2.3.0.

TASKS.md

@@ -1,6 +1,16 @@
 # derp - Tasks
 
-## Current Sprint -- v2.2.0 Configurable Proxy (2026-02-21)
+## Current Sprint -- v2.3.0 Mumble Music Playback (2026-02-21)
+
+| Pri | Status | Task |
+|-----|--------|------|
+| P0 | [x] | `src/derp/opus.py` -- ctypes libopus encoder (zero Python deps) |
+| P0 | [x] | `src/derp/mumble.py` -- voice varint, voice packet, PCM scaling, stream_audio |
+| P0 | [x] | `plugins/music.py` -- play/stop/skip/queue/np/volume commands |
+| P1 | [x] | Tests: `test_opus.py`, `test_mumble.py` voice additions, `test_music.py` |
+| P2 | [x] | Documentation update (USAGE.md, CHEATSHEET.md) |
+
+## Previous Sprint -- v2.2.0 Configurable Proxy (2026-02-21)
 
 | Pri | Status | Task |
 |-----|--------|------|

docs/CHEATSHEET.md

@@ -540,10 +540,27 @@ operators = []
 trusted = []
 ```
 
-TCP/TLS via SOCKS5 proxy by default (`proxy = true`). Text chat only (no voice). Minimal protobuf
+TCP/TLS via SOCKS5 proxy by default (`proxy = true`). Minimal protobuf
 codec (no external dep). HTML stripped on receive, escaped on send.
 IRC-only commands are no-ops. ~90% of plugins work.
 
+## Music (Mumble only)
+
+```
+!play <url>              # Play audio (YouTube, SoundCloud, etc.)
+!stop                    # Stop playback, clear queue
+!skip                    # Skip current track
+!queue                   # Show queue
+!queue <url>             # Add to queue (alias for !play)
+!np                      # Now playing
+!volume                  # Show current volume
+!volume 75               # Set volume (0-100, default 50)
+```
+
+Requires: `yt-dlp`, `ffmpeg`, `libopus.so.0` on the host.
+Max 50 tracks in queue. Volume applies on next track.
+Mumble-only: `!play` replies with error on other adapters, others silently no-op.
+
 ## Plugin Template
 
 ```python

docs/USAGE.md

@@ -1553,6 +1553,36 @@ Mumble uses HTML for text messages. On receive, the bot strips tags and
 unescapes entities. On send, text is HTML-escaped. Action messages use
 `<i>` tags for italic formatting.
 
+### Music Playback
+
+Stream audio from YouTube, SoundCloud, and other yt-dlp-supported sites
+into the Mumble voice channel. Audio is decoded to PCM, encoded to Opus
+via system libopus, and transmitted as voice packets over the TCP tunnel.
+
+**System dependencies** (must be installed on the host):
+- `yt-dlp` -- audio stream extraction
+- `ffmpeg` -- decode to 48kHz mono s16le PCM
+- `libopus.so.0` -- Opus codec (ctypes, no Python binding)
+
+```
+!play <url>          Play audio or add to queue
+!stop                Stop playback, clear queue
+!skip                Skip current track
+!queue               Show queue
+!queue <url>         Add to queue (alias for !play)
+!np                  Now playing
+!volume [0-100]      Get/set volume
+```
+
+- Queue holds up to 50 tracks
+- Volume applies from the next track (default: 50%)
+- Title resolved via `yt-dlp --get-title` before playback
+- Audio pipeline: `yt-dlp | ffmpeg` subprocess, 20ms Opus frames
+- Commands are Mumble-only; `!play` on other adapters replies with an error,
+  other music commands silently no-op
+- Playback runs as an asyncio background task; the bot remains responsive
+  to text commands during streaming
+
 ### Transport
 
 TCP connections route through the SOCKS5 proxy at `127.0.0.1:1080`