The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted article titles and summaries from external web searches and returns them to the agent. This constitutes an indirect prompt injection surface if the source content contains malicious instructions designed to manipulate the LLM's behavior.
Ingestion points: scripts/search_wechat.js performs HTTPS requests to Sogou and WeChat domains to retrieve article metadata.
Boundary markers: The results are returned as a structured JSON object, providing a data boundary for the agent's context.
Capability inventory: The skill has network access (HTTPS) and local file writing capabilities (via the -o CLI argument).
Sanitization: The script parses and trims the content but does not perform specialized sanitization to prevent LLM prompt injection attacks.
[EXTERNAL_DOWNLOADS]: The skill requires the cheerio package, which is a well-known, established library for HTML manipulation in the Node.js ecosystem. The installation instructions follow standard package management practices.

wechat-article-search