favstash-shortform

Pass

Audited by Gen Agent Trust Hub on Sep 11, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill frequently executes external command-line tools to perform its core functions. It uses ffmpeg and ffprobe for analyzing video properties, sampling frames, and transcoding audio. It also uses yt-dlp for media retrieval and npm for workspace setup.
  • [EXTERNAL_DOWNLOADS]: The scripts/reference-analyze.mjs utility uses yt-dlp to download video files and metadata from user-provided URLs. The script also includes an optional --cookies-from-browser flag which, while a standard yt-dlp feature for accessing authenticated content, could theoretically be used to access local browser cookies if misconfigured by a user.
  • [EXTERNAL_DOWNLOADS]: The scripts/init-workspace.mjs script performs an npm install of the hyperframes package (version 0.8.10) to set up a local rendering environment. This involves fetching and installing third-party code from the npm registry.
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process data from external sources, creating a potential vector for indirect prompt injection.
  • Ingestion points: Remote video content and associated metadata downloaded via yt-dlp; user-provided creator preferences stored in preferences.md.
  • Boundary markers: The skill relies on descriptive instructions to guide the agent but does not implement strict technical boundary markers (like specific delimiters) for the data it processes.
  • Capability inventory: The agent has the ability to execute shell commands, perform file system operations, and communicate with external URLs (via tools).
  • Sanitization: The skill employs path sanitization and directory assertion (assertInside) to prevent path traversal, and uses yt-dlp's built-in filename restriction features.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 11, 2026, 11:30 AM