The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted content from X/Twitter posts via the FxTwitter API. It addresses this risk by implementing an escapeHtml function in scripts/yichen_x_slicer.mjs to sanitize text before it is rendered in a browser environment. It also validates that post IDs are strictly numeric.
Ingestion points: readSource in scripts/yichen_x_slicer.mjs fetches JSON data from api.fxtwitter.com.
Boundary markers: No specific delimiters are used for the processed text, but the data is handled as structured JSON and escaped before insertion into the DOM.
Capability inventory: The skill can execute local binaries (ffmpeg, ffprobe, zip, unzip) and launch playwright for rendering.
Sanitization: escapeHtml is applied consistently to all tweet text and metadata during the generation of the HTML frames.
[EXTERNAL_DOWNLOADS]: The skill downloads media assets required for processing. It restricts these downloads to a whitelist of trusted domains (pbs.twimg.com, video.twimg.com, and api.fxtwitter.com) and incorporates signature verification to ensure downloaded files are valid image or MP4 formats before they are accessed by system tools.
[COMMAND_EXECUTION]: Local shell utilities such as ffmpeg and zip are used via spawnSync. The arguments are strictly constructed from validated inputs, and the skill includes a safeOutputPath utility to prevent directory traversal by ensuring all file writes are contained within the designated output directory.

yichen-x-slicer