ai-podcast-creation
Pass
Audited by Gen Agent Trust Hub on Mar 7, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill suggests installing external packages from 'inference-sh/skills' via the 'npx' package runner to enable broader agent capabilities.- [COMMAND_EXECUTION]: All primary operations are conducted through the 'infsh' command-line tool, which interacts with the remote inference.sh infrastructure for audio processing and LLM inference.- [PROMPT_INJECTION]: The 'NotebookLM-Style' feature creates a surface for indirect prompt injection. * Ingestion points: User-provided document content is directly inserted into a generation prompt for the Claude model within 'SKILL.md'. * Boundary markers: No delimiters or 'ignore' instructions are used to separate user data from the system prompt, which may lead to the agent following instructions embedded in the document. * Capability inventory: Influenced outputs can be passed to subsequent tools like 'media-merger' or 'kokoro-tts' via 'infsh app run'. * Sanitization: The skill does not perform any sanitization or validation on the text extracted from the source documents before processing.
Audit Metadata