speak

Warn

Audited by Snyk on Mar 2, 2026

Risk Level: MEDIUM
Full Analysis

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

  • Third-party content exposure detected (high risk: 0.70). The skill explicitly downloads and consumes arbitrary HTTP(S) reference audio (see render_timeline.py's _resolve_reference_audio and the SKILL.md/ noiz examples that allow --ref-audio with a URL), so untrusted third‑party audio is ingested and used to control voice‑cloning/TTS output and thus can materially influence behavior.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

  • Potentially malicious external URL detected (high risk: 0.80). The skill calls the Noiz API (base URL https://noiz.ai/v1) at runtime (when using the Noiz backend / --auto-emotion) and its /emotion-enhance response is injected back as the text used for synthesis, so remote content directly controls the prompts/inputs and the Noiz service is a required dependency for that backend.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Mar 2, 2026, 02:22 PM