The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute several shell commands, including wc for word counting, pdflatex and bibtex for compiling documents, and convert or pdftoppm for processing PDF pages into images.\n- [DYNAMIC_EXECUTION]: The instructions include a specific Python code snippet to be executed via python -c which performs document processing using the pdf2image library.\n- [EXTERNAL_DOWNLOADS]: The skill performs runtime installation of the pdf2image Python package using uv add and retrieves external data through research tools like parallel-web and research-lookup.\n- [PROMPT_INJECTION]: The instructions include directives aimed at overriding the agent's default operational boundaries, such as 'ALWAYS complete the ENTIRE task without stopping', 'NEVER ask "Would you like me to continue?"', and 'Token usage is unlimited'.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted research data from external websites and academic papers, presenting an injection surface.\n
Ingestion points: Web search results and academic paper metadata/content retrieved via external tools.\n
Boundary markers: None provided; the instructions do not include delimiters or specific guidance to ignore embedded instructions in external data.\n
Capability inventory: The skill has access to the local file system, shell command execution, Python runtime, and various specialized analysis tools.\n
Sanitization: No evidence of content sanitization or filtering before external data is incorporated into documents or saved to disk.

writing