ai-podcast-creation

Pass

Audited by Gen Agent Trust Hub on Mar 18, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [SAFE]: The skill facilitates the generation of audio content, music, and merged podcast episodes using the author's own infrastructure and established third-party services like OpenRouter.
  • [COMMAND_EXECUTION]: The skill utilizes the 'infsh' command to execute various tasks. Access is restricted via the 'allowed-tools' metadata to 'Bash(infsh *)', ensuring the agent cannot execute arbitrary shell commands outside of the intended CLI interface.
  • [EXTERNAL_DOWNLOADS]: The documentation provides a link to an installation script hosted on the official GitHub repository for 'inference-sh'. As a vendor-owned resource, this is considered a safe reference for tool setup.
  • [PROMPT_INJECTION]: The skill includes features to generate conversational scripts from documents provided by users.
  • Ingestion points: Untrusted document content is passed to the Claude model via the 'infsh app run openrouter/claude-sonnet-45' command in SKILL.md.
  • Boundary markers: The example prompts do not currently utilize explicit delimiters (like XML tags or triple quotes) to isolate user document content from the instructions.
  • Capability inventory: The skill possesses the ability to generate and store JSON files, and trigger audio synthesis tools via 'infsh'.
  • Sanitization: No explicit sanitization or filtering of the document content is defined within the provided skill logic.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 18, 2026, 08:37 AM