ASR
Warn
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: MEDIUMEXTERNAL_DOWNLOADS
Full Analysis
- External Downloads (MEDIUM): The script relies on 'z-ai-web-dev-sdk', an external Node.js package from a source not included in the trusted organizations or repositories lists. The behavior of this SDK and its network communication targets cannot be verified through static analysis of this skill.
- Indirect Prompt Injection (LOW): The skill ingests audio data which is transcribed into text. This transcription process represents a vulnerability surface where instructions spoken in the audio could influence the agent's behavior. Ingestion points: Read of './output.wav' in scripts/asr.ts. Boundary markers: None. Capability inventory: File system access (fs.readFileSync) and network access (via SDK). Sanitization: No validation or sanitization is performed on the transcription result.
Audit Metadata