The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill's primary purpose is to process untrusted PDF documents provided by users, which creates a surface for indirect prompt injection attacks where malicious instructions hidden in the PDF could influence the agent's subsequent actions.
Ingestion points: PDF files are parsed using libraries like pdfplumber, pypdf, and pypdfium2 as detailed in SKILL.md and docs/reference.md.
Boundary markers: The instructions do not recommend using specific delimiters or "ignore embedded instructions" warnings when the agent handles extracted content.
Capability inventory: The skill possesses file system access for document manipulation and the capability to execute Python scripts.
Sanitization: No sanitization or validation steps are defined for the extracted text or table data before the agent processes it.
[COMMAND_EXECUTION]: The skill relies on shell commands for environment setup and task execution, which is documented as standard behavior for its intended use case.
Evidence: docs/pdf_to_word_conversion_guide.md provides commands for package management (pip3 install pdf2docx) and script execution (python3 convert.py).
[DYNAMIC_EXECUTION]: The skill workflow involves generating and executing Python scripts at runtime to handle specific document transformations.
Evidence: The conversion guide in docs/pdf_to_word_conversion_guide.md instructs the agent to create a temporary convert.py script from a provided template and execute it to perform PDF to Word conversion.
[EXTERNAL_DOWNLOADS]: The skill extends its capabilities by installing well-known libraries from official registries.
Evidence: The requirements.txt file and conversion documentation list dependencies including pypdf, pdfplumber, reportlab, pandas, openpyxl, and pdf2docx.

pdf