podcast-downloader
Pass
Audited by Gen Agent Trust Hub on Oct 5, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external sources (iTunes API and RSS feeds) and incorporates it into local files and filenames, which could potentially influence an AI agent's behavior if it reads these files back.
- Ingestion points: External data is fetched from the iTunes API and RSS feeds via the
requestsandfeedparserlibraries inscripts/download_podcast.py. - Boundary markers: The skill does not provide specific instructions or delimiters to the agent to treat the downloaded metadata (titles, descriptions) as untrusted content.
- Capability inventory: The skill possesses the capability to write to the file system (saving audio and JSON metadata) and perform network operations (fetching API data and audio streams).
- Sanitization: The
sanitize_filenamefunction is used to clean episode titles before they are used as filenames, mitigating basic path traversal risks, though the metadata fields themselves are saved as-is in JSON format. - [COMMAND_EXECUTION]: The skill involves the execution of a Python script to perform its core functions. The implementation is focused on the intended task and does not show signs of arbitrary command injection or dangerous subprocess usage.
- [EXTERNAL_DOWNLOADS]: The skill queries metadata from Apple's iTunes API (a well-known service) and downloads audio files from external URLs provided within the podcast metadata.
Audit Metadata