record-web-journey

Pass

Audited by Gen Agent Trust Hub on Sep 5, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The setup.sh script downloads Node.js packages from the NPM registry and fetches browser binaries (Chromium, ffmpeg) using Playwright. It also uses the agent-browser tool to fetch additional components. All downloads target established registries or the framework's own components.
  • [COMMAND_EXECUTION]: The skill uses shell commands to inspect the host environment (e.g., checking for X11 displays with pgrep and shared memory with df) and to manage the browser process. These actions are consistent with its primary purpose of recording web sessions.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes and logs content from external websites, creating a surface for indirect prompt injection.
  • Ingestion points: Captures titles, URLs, and visual representations from any website the agent or user visits.
  • Boundary markers: The output log steps.md does not employ specific delimiters to isolate potentially malicious text found on external pages.
  • Capability inventory: The skill possesses the ability to write to the local file system and execute browser subprocesses.
  • Sanitization: It redacts credentials from captured URLs and avoids logging typed values, but other page content is stored without sanitization.
  • [DYNAMIC_EXECUTION]: The recorder script utilizes page.evaluate and addInitScript to execute JavaScript within the browser context. This is used to place visual markers at click locations and monitor navigation events for the recording.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 5, 2026, 01:54 AM
Security Audit — agent-trust-hub — record-web-journey