diff --git a/listen/SKILL.md b/listen/SKILL.md
new file mode 100644
index 000000000..cf0d0d355
--- /dev/null
+++ b/listen/SKILL.md
@@ -0,0 +1,42 @@
+# /listen — System Audio Capture
+
+Capture system audio from Linux environments using PipeWire or PulseAudio. Auto-detects the active audio backend and records to WAV.
+
+gstack has eyes (`/browse`). This gives it ears.
+
+## Usage
+
+```
+/listen [duration_seconds] [output_file]
+```
+
+**Examples:**
+```
+/listen              # 5 seconds -> captured.wav
+/listen 10           # 10 seconds -> captured.wav
+/listen 5 meeting.wav
+```
+
+## How it works
+
+1. Detects audio backend: PipeWire (`pw-record`) or PulseAudio (`parec`)
+2. For PulseAudio: auto-discovers the monitor source
+3. Captures system audio for the specified duration
+4. Saves as WAV file
+
+## Requirements
+
+- Linux with PipeWire or PulseAudio
+- `pw-record` (PipeWire) or `parec` (PulseAudio)
+- Python 3.6+
+
+## Use cases
+
+- ** app QA**: Capture app audio output for automated verification
+- **Meeting capture**: Record system audio during calls
+- **CI audio testing**: Validate audio output in automated pipelines
+
+## Tested on
+
+- Ubuntu 24.04 LTS — PipeWire 1.0.5 (real hardware)
+- WSL2 Ubuntu 24.04 — PulseAudio (WSLg)