stitched-full-page-capture

Pass

Audited by Gen Agent Trust Hub on Aug 26, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The script scripts/stitch_full_page_capture.mjs executes the ffmpeg and sips binaries using node:child_process.spawn and node:child_process.execFile. These tools are used for image processing (cropping, stitching) and metadata retrieval, respectively. The script constructs command arguments as arrays, preventing shell injection.- [PROMPT_INJECTION]: The skill processes external web content by navigating to URLs provided in a manifest file. This ingestion of untrusted data constitutes an indirect prompt injection attack surface, although the risk is minimized as the skill's output is primarily visual (images) rather than text that would be interpreted as instructions by the agent.
  • Ingestion points: External web pages accessed via pageUrl in the manifest.
  • Boundary markers: None; navigation is direct to the untrusted source.
  • Capability inventory: File system writes (fs.writeFile, fs.mkdir), subprocess execution (ffmpeg, sips), and network navigation (Playwright).
  • Sanitization: Item titles are sanitized using character replacement before being used in file paths, and browser navigation is isolated within a headless Chromium instance.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 26, 2026, 01:37 PM
Security Audit — agent-trust-hub — stitched-full-page-capture