voice-ui

SUSPICIOUS. The core idea is coherent for a voice UI skill, but the footprint is broader than a simple UI helper: it can rewrite local code, auto-commit changes, and may read OpenClaw config for credentials. No direct malicious exfiltration is evident, yet the unseen start script/dependencies and autonomous self-modification make this a medium-risk skill.

No clear evidence of intentional malware (no backdoor/persistence/command execution) in this client fragment. However, it has major security risks: (1) DOM XSS because message text from transcription/chat is rendered with innerHTML without proper escaping, and (2) sensitive API key exposure by fetching /api/key and using it directly in browser Authorization headers, enabling third-party API abuse if the key is leaked or an attacker gains script execution. Overall, security should be reviewed and remediated before use.

High security risk. This module (1) explicitly discloses the OPENAI_API_KEY to any caller via GET /api/key, (2) likely enables unauthorized file reads by using req.url-derived paths for fs.readFileSync without strong traversal protection, and (3) forwards user-controlled message content to a local OpenClaw executable and returns its output to the client. While shell injection is mitigated by using spawn with args, the overall trust boundaries are dangerous and could enable data theft or unauthorized access depending on the local agent’s capabilities.