The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/render.py uses subprocess.run() to execute ffmpeg for video encoding. This is an expected part of the skill's functionality to convert PNG frames into a video file. The command uses parameters derived from the local config.json (such as frame rate) and inputs strictly controlled by the script's previous steps.
[REMOTE_CODE_EXECUTION]: Static analysis identified the use of Playwright to drive a headless browser. The browser navigates to a locally generated HTML file (hyperframe.html) and executes a deterministic JavaScript function (window.renderAt(t)) to step through frames. No external URLs are visited, and no untrusted remote code is executed.
[DATA_EXPOSURE]: The skill reads from a local config.json and local asset paths (images/SVGs). It does not access sensitive system files, environment variables, or credentials. All file operations are scoped to the current working directory or paths specified in the configuration.
[PROMPT_INJECTION]: The SKILL.md file contains standard instructions for the agent to follow a specific rendering format. There are no patterns suggesting attempts to override safety filters or hijack agent behavior.

render-mosaic-grid-reveal