The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes external system binaries including Chrome/Chromium and FFmpeg using the subprocess module to handle video rendering and frame extraction tasks.
Evidence: scripts/render.py and scripts/check_frames.py invoke ffmpeg via CLI arguments to process generated video files.
Evidence: scripts/livepanel.py launches Chrome using the --remote-debugging-pipe flag to enable headless automation.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied JSON configuration files which could theoretically contain malicious instructions designed to influence the agent's interpretation of the system topology.
Ingestion points: scripts/render.py and scripts/check_frames.py ingest local JSON data via the --config argument.
Boundary markers: SKILL.md explicitly warns the agent to treat topology as data and not instructions, and the HTML template implements a restrictive CSP to prevent external resource loading.
Capability inventory: The skill has local file system access and the ability to trigger subprocesses.
Sanitization: The renderer (assets/template.html) escapes text content before rendering to mitigate HTML injection risks.
[DYNAMIC_EXECUTION]: The skill utilizes the Chrome DevTools Protocol to execute dynamic JavaScript within a headless browser to manage the rendering lifecycle and perform layout validation.
Evidence: scripts/livepanel.py implements a Chrome.eval() method that interfaces with the page context via Runtime.evaluate calls.
Evidence: scripts/check_frames.py constructs JavaScript strings to perform pixel-level comparisons of captured PNG frames within the browser's canvas environment.

live-panel