The Agent Skills Directory

[COMMAND_EXECUTION]: The scripts/integrate_screenshot.py script uses subprocess.run to invoke a local Chrome or Chromium executable. This execution is used specifically for the verify and self-test subcommands to render the modified HTML in a headless browser and confirm that the screenshot dimensions match the expected output. The command uses the file:// protocol and operates on temporary local files.
[PERSISTENCE]: The skill includes scripts/profile_store.py which manages a persistent user profile in the user's home directory (e.g., ~/.config/agent-skills/profile.json). This is used to store non-sensitive preferences like default_scale or default_width across different agent sessions. The script includes an explicit blacklist for sensitive keywords such as 'token', 'secret', or 'api_key' to prevent accidental exposure of credentials.
[INDIRECT_PROMPT_INJECTION]: The skill processes external HTML files provided by the user. While this creates an ingestion point for untrusted data, the skill performs modifications using deterministic Python logic (regex-based element wrapping and script injection) rather than passing the content to the LLM for instruction extraction. The capability to write to the local filesystem is limited to the target HTML file specified by the user.

lov-integrate-modern-screenshot