The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/export_tutorial.py uses subprocess.run to invoke local system tools, specifically pandoc for document conversion and headless browsers (Chrome, Edge, etc.) for PDF generation. The execution is handled securely using list-based arguments without spawning a shell, preventing common command injection vulnerabilities.
[EXTERNAL_DOWNLOADS]: The skill references documentation and guidance from well-known technology sites such as pandoc.org and readthedocs.io, as well as public GitHub repositories for technical benchmarks. These references are used to guide the agent's research and formatting workflows.
[DATA_EXFILTRATION]: As part of its primary function, the skill performs web research to gather information from authoritative sources. This involves making network requests to external domains to fetch learning content.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from the web (research data) and interpolates it into a tutorial format.
Ingestion points: External source URLs gathered during the research phase defined in references/authority-research.md.
Boundary markers: The skill defines a 'Source Appendix' to separate citations, but does not use explicit boundary markers to wrap untrusted content during processing.
Capability inventory: Uses subprocess.run in scripts/export_tutorial.py to call conversion tools.
Sanitization: Relies on the standard parsing and rendering behavior of Pandoc and web browsers rather than custom sanitization logic.

learning-builder