The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests thread data and brand configuration from a JSON file to generate video content. This data is interpolated into HTML templates that are then rendered by a local browser instance. While the skill uses an escapeHTML helper for text bubbles, other fields such as logo_svg permit the inclusion of raw SVG data, providing a potential surface for injecting scripts into the rendering context.
Ingestion points: config.json (thread messages and participants, end card configuration) and local file system paths for images and SVGs.
Boundary markers: None provided to distinguish instructions from data within the interpolated content.
Capability inventory: node, playwright (Chromium), ffmpeg, ffprobe, and python3 (subprocess execution).
Sanitization: escapeHTML is applied to text bubbles and CTA text, but not to raw SVG inputs or file path parameters.
[COMMAND_EXECUTION]: The skill invokes several command-line utilities, including ffmpeg and ffprobe, to process video clips and stitch them together. Parameters for these commands are derived from the input configuration, though they are handled with shell quoting in the provided scripts.
[EXTERNAL_DOWNLOADS]: The skill fetches the Playwright library and Chromium browser from established registries to facilitate its rendering capabilities.
[DYNAMIC_EXECUTION]: The skill performs runtime assembly of HTML content and executes an inline Python helper script to manage deterministic audio mixing and SFX cueing.

render-imessage-chat