assemblyai-common-errors

Pass

Audited by Gen Agent Trust Hub on Sep 12, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes data from external API responses and webhooks, creating a potential attack surface for indirect prompt injection. However, it provides extensive safety instructions to mitigate this risk.
  • Ingestion points: Triages REST, transcript, and streaming responses as defined in SKILL.md.
  • Boundary markers: Explicit instructions to "Exclude credentials, signed URLs, audio, transcript text, prompts, and customer-derived content" and "never reproduce speech content" serve as strong guidance for the agent.
  • Capability inventory: Tool access is restricted in the frontmatter to Read, Glob, Grep, Write, and Edit.
  • Sanitization: Instructions mandate collecting only "safe status, code, timing, and request metadata" to prevent data leakage.- [EXTERNAL_DOWNLOADS]: The skill references external resources from the official AssemblyAI domain.
  • The references/official-docs.md file links to official API documentation, machine-readable guides, and migration instructions on assemblyai.com, which is a well-known service provider.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 12, 2026, 04:49 AM
Security Audit — agent-trust-hub — assemblyai-common-errors