The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill fetches content from untrusted external URLs (WeChat articles). This content is processed and saved as Markdown, creating a potential surface for indirect prompt injection if the resulting file is later processed by another AI agent.
Ingestion points: scripts/fetch.js fetches HTML content from user-supplied URLs via axios.
Boundary markers: The generated Markdown includes Front Matter, but does not include explicit warnings to downstream agents to ignore instructions within the article body.
Capability inventory: The skill writes files (Markdown and images) to the local file system in a temporary directory.
Sanitization: Basic character replacement is performed for filenames and YAML Front Matter metadata, but the article content itself is not sanitized for malicious instructions.
[EXTERNAL_DOWNLOADS]: The skill downloads image assets from external servers during the article processing phase.
Evidence: scripts/fetch.js iterates through image elements and fetches them using axios.get(imgUrl, ...) to save them locally in the attachments folder.

wechat-article-fetcher