ag-testlisted

Push a test frame to the SmallTV so the user can check the display. ## Preflight ```bash echo "env AGENT_GLANCE_IP=${AGENT_GLANCE_IP:-<unset>}" cat ~/.agent-glance/config.json 2>/dev/null || echo "(no saved config)" ``` If neither is set, the device has never been onboarded — stop and tell the user to run the `ag-setup` skill first. Do not guess an IP. ## Push Take the state (`working` | `waiting` | `done`) and an optional subtitle from what the user asked for. Default to `waiting` (the most visually distinct) and `test frame` when they did not say. ```bash python3 "${CLAUDE_PLUGIN_ROOT}/scripts/agent_glance.py" --test "<state>" "<subtitle>" ``` If the user did not name a state, cycle all three with a pause so they can watch the transitions, ending on `done`: ```bash for s in working waiting done; do python3 "${CLAUDE_PLUGIN_ROOT}/scripts/agent_glance.py" --test "$s" "render check" sleep 3 done ``` ## Confirm ```bash [ -s /tmp/agent_glance_error.log ] && tail -5 /tmp/agent_glance_error.log || echo "no errors" ``` Expected: amber WORKING, red APPROVAL, green DONE — each with the model name, a context-usage bar, and in/out token counts along the bottom. `--test` respects the current preset: in gif mode (`--preset hosts`/`custom`) the middle of the frame is the looping character GIF instead of the status text, and the device animates it. State still shows via the top accent bar + background colour. If gif mode is on but no character resolves, it falls back to the