The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill employs dynamic script generation in scripts/lib/launch-check.mjs. It creates a JavaScript file (entry.mjs) in a temporary system directory and executes it using ego-browser nodejs -e "await import(...)". This allows for the assembly and execution of code at runtime.
[COMMAND_EXECUTION]: The skill uses child_process.spawn and spawnSync to execute the ego-browser binary and node scripts. This is the primary mechanism for its login checks and file serving utilities.
[DATA_EXPOSURE]: The script scripts/serve-file.mjs launches a local HTTP server that serves a user-defined file on 127.0.0.1. The server is configured with Access-Control-Allow-Origin: *, which enables the browser to fetch the file for upload but also exposes the file's content to other local processes or browser tabs that can reach the local port.
[INDIRECT_PROMPT_INJECTION]: The skill acts as a surface for indirect prompt injection by processing untrusted user content for video metadata.
Ingestion points: User-supplied strings for description, tags, and short_title are defined in references/inputs.md and processed in SKILL.md.
Boundary markers: No boundary markers or instructions to ignore embedded commands are present when these strings are injected into the browser context via page.evaluate and document.execCommand.
Capability inventory: The skill possesses file system access (via node and serve-file.mjs) and browser automation capabilities through the ego-browser wrapper.
Sanitization: Input sanitization is limited to WeChat platform compliance (e.g., character length and symbol whitelists for titles) rather than security-focused filtering.

wechat-channels-video