youtube-digest
Pass
Audited by Gen Agent Trust Hub on Sep 17, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses shell scripts (
scripts/extract_metadata.shandscripts/extract_transcript.sh) to invoke theyt-dlpcommand-line utility. While the scripts properly quote the URL variable ("$URL") to mitigate basic command injection, the execution of external tools on untrusted URLs always carries a marginal risk if the tool itself has parsing vulnerabilities. - [INDIRECT_PROMPT_INJECTION]: The skill's primary function involves ingesting and processing untrusted data from external sources, which could contain malicious instructions designed to influence the agent's behavior.
- Ingestion points: Video transcripts and metadata are fetched using
yt-dlp. Additionally, the "Deep Research" workflow (references/deep-research.md) usesWebSearchandWebFetchto gather information from arbitrary websites. - Boundary markers: The provided documentation does not specify the use of clear delimiters or instructions to ignore embedded commands within the fetched content, increasing the risk that the agent might follow instructions hidden in transcripts or web pages.
- Capability inventory: The skill possesses the ability to execute shell scripts and write files to the local file system (
research/readings/youtube/). - Sanitization: There is no evidence of sanitization, filtering, or validation of the content retrieved from YouTube or the web before it is interpolated into the prompts for summary and quiz generation.
Audit Metadata