The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill uses curl to fetch content from r.jina.ai, which is a well-known service for converting web pages into LLM-friendly formats. It also installs common Python libraries (markdownify, requests, beautifulsoup4) using pip install.
[COMMAND_EXECUTION]: The skill executes a Python script block using a here-document (`python
<<'PY') to process HTML content locally. While the script logic is transparently defined in the skill, it involves file system writes (open("article.md","w").write(out)`).
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from arbitrary external URLs.
Ingestion points: Data enters the context via curl or requests.get from user-supplied URLs in SKILL.md.
Boundary markers: The output is written directly to a file and read by the agent without explicit delimiters or instructions to ignore embedded commands.
Capability inventory: The skill possesses capabilities for file writing and package installation via pip.
Sanitization: The Python script cleans HTML tags (script, style, nav, etc.) using BeautifulSoup, but does not sanitize the text content for potential adversarial instructions targeting the agent.

web-to-markdown