boring-youtube-mining

Pass

Audited by Gen Agent Trust Hub on Mar 9, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill invokes system commands including yt-dlp, mkdir, and ls. It interpolates variables like {handle} and {video_id} directly from local configuration files (settings/youtube-channels.md) and tool output into shell command strings without explicit sanitization, which could lead to command injection if the input data is maliciously crafted.
  • [EXTERNAL_DOWNLOADS]: The skill fetches video metadata and auto-generated transcripts from YouTube's official domain using the yt-dlp utility. These downloads are part of the skill's primary function and target a well-known service.
  • [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it ingests untrusted transcript data from YouTube to generate user-facing content ideas and draft hooks.
  • Ingestion points: YouTube transcripts are downloaded and stored in the content/youtube-transcripts/ directory for analysis.
  • Boundary markers: The skill does not implement delimiters or specific instructions to the agent to treat the transcript content as untrusted data or to ignore instructions embedded within the text.
  • Capability inventory: The skill executes subprocesses for yt-dlp (including network access), creates directories, and writes files to the local filesystem.
  • Sanitization: Sanitization is limited to structural cleaning of the VTT format (removing timestamps and metadata); no semantic sanitization or filtering of the transcript text is performed to mitigate injection attacks.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 9, 2026, 09:23 PM