ai-podcast-creation
Pass
Audited by Gen Agent Trust Hub on Sep 21, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill relies extensively on the
beltCLI tool viaBash(belt *)to interact with external AI models for text-to-speech, music generation, and audio editing. - [EXTERNAL_DOWNLOADS]: The skill references and encourages the download of external components, specifically the
beltCLI (npx skills add belt-sh/cli) and installation guidelines from GitHub (raw.githubusercontent.com/inference-sh/skills/...). These are central to the skill's intended functionality on theinference.shplatform. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data (documents and text) by interpolating them into prompts for LLMs and TTS models.
- Ingestion points: External document content (
<your-document-content>) and script lines are ingested via prompt templates inSKILL.md. - Boundary markers: The templates lack explicit boundary markers or instructions for the agent to ignore potentially malicious instructions embedded within the processed text.
- Capability inventory: The skill uses
Bash(belt *)to run remote applications for LLM processing, audio synthesis, and media merging. - Sanitization: There is no evidence of sanitization or validation of the input data before it is interpolated into the CLI command arguments.
Audit Metadata