Walkie-Talkie Mode

Handles voice-to-voice conversations on WhatsApp. Automatically transcribes incoming audio and responds with local TTS audio. Use when the user wants to "talk" instead of type.

This skill's behavior (transcribe incoming audio, produce local TTS, send .ogg back) matches its description, but the SKILL.md depends on local tools and scripts that are not declared anywhere. Before installing or enabling: 1) Verify the agent environment actually has the required binaries (ffmpeg, whisper-cpp, sherpa-onnx-tts) and the helper script paths (tools/transcribe_voice.sh, bin/sherpa-onnx-tts). 2) Ask the author to update metadata to list required binaries, exact paths, and any model files or hardware needs. 3) Confirm the 'message' tool used to send files is the authorized platform tool (so audio is sent only to the intended chat) and that no unexpected external endpoints are contacted. 4) Review file permissions around /tmp and any model data to avoid exposing unrelated data. 5) Test in a sandboxed agent first — if the required local tools are missing, the skill will fail or may attempt to run arbitrary local programs if created later. If you cannot verify or supply the declared dependencies, treat this skill as untrusted.

Type: OpenClaw Skill Name: walkie-talkie-mode Version: 1.0.0 The skill bundle describes a 'walkie-talkie' mode for voice-to-voice conversations, primarily using local tools for transcription and text-to-speech. The `SKILL.md` file clearly outlines the workflow, triggers, and constraints, instructing the agent to execute local scripts (`tools/transcribe_voice.sh`) and binaries (`bin/sherpa-onnx-tts`) for its stated purpose. There is no evidence of prompt injection attempting to subvert the agent, exfiltrate data, establish persistence, or perform other malicious actions. The instructions are straightforward and align with the benign functionality described.