ai-podcast-creation

Pass

Audited by Gen Agent Trust Hub on Sep 2, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill features workflows that process user-provided documents to generate podcast scripts using LLMs, which creates a surface for potential instruction injection from the source material. However, this is inherent to the skill's primary purpose. \n
  • Ingestion points: The \u003cyour-document-content\u003e placeholder used in the belt app run openrouter/claude-sonnet-45 command within SKILL.md.\n
  • Boundary markers: No explicit delimiting markers or ignore-instructions are used for the document content in the examples.\n
  • Capability inventory: The skill is restricted via allowed-tools to the belt CLI, which interacts with audio processing and AI models.\n
  • Sanitization: There is no evidence of sanitization or filtering for the ingested text.\n- [SAFE]: No malicious patterns such as credential exfiltration, persistence mechanisms, or obfuscated code were detected. The skill correctly scopes its execution environment using the allowed-tools frontmatter and points to legitimate service documentation.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 2, 2026, 02:23 PM
Security Audit — agent-trust-hub — ai-podcast-creation