ai-voice-chat

[Skill Scanner] Installation of third-party script detected This skill's declared purpose matches its requested capabilities: local STT/TTS, a local gateway, and auto-start on headset. Sensitive items are required (OpenClaw API token and a session target). The manifest takes reasonable precautions (Keychain, localhost-only API), but several supply-chain and operational risks remain: the setup script downloads models (download-and-execute pattern), the skill auto-runs in the background, and it delegates transcript forwarding and real-world actions to a local gateway whose integrity is critical. If OpenClaw or the local machine is compromised the skill could leak transcripts, tokens, or initiate actions to external parties. Overall the module is NOT overtly malicious, but it carries a meaningful supply-chain and privilege risk and should be audited before use (inspect setup.sh and runtime scripts, verify model download sources and checksums, and confirm localhost-only enforcement in code). LLM verification: Documentation describes a legitimate-sounding local voice assistant skill whose capabilities align with its stated purpose. The primary security concern is supply-chain: the setup/install step (brew and scripts/setup.sh) and model downloads are remote operations with no shown verification, and the actual scripts that enforce the 'localhost-only' policy were not provided for review. These factors raise a moderate security risk that must be mitigated by reviewing the setup script and any installed