descript-api

Pass

Audited by Gen Agent Trust Hub on Mar 14, 2026

Risk Level: SAFECREDENTIALS_UNSAFEEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [CREDENTIALS_UNSAFE]: The skill explicitly documents the locations of sensitive credentials, identifying that the DESCRIPT_API_KEY is stored in ~/.zshrc and ~/.descript-cli/config.json. Highlighting these paths may allow an agent to access or expose authentication secrets.
  • [EXTERNAL_DOWNLOADS]: The skill includes instructions to download and install the @descript/platform-cli package via npm. This is a well-known package provided by the official service provider.
  • [PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection by processing external data to influence subsequent AI actions.
  • Ingestion points: External transcript data is fetched via GET /published_projects/{slug} (documented in SKILL.md).
  • Boundary markers: None identified; external transcript content is parsed and potentially interpolated directly into instructions.
  • Capability inventory: The skill uses ingested data to craft natural language prompts for the 'Underlord' agent via POST /jobs/agent (documented in SKILL.md).
  • Sanitization: No specific sanitization or filtering of the transcript content for malicious instructions is implemented beyond basic VTT parsing.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 14, 2026, 07:07 PM