The Agent Skills Directory

[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection due to its core function of processing untrusted external content.\n
Ingestion points: scripts/extract_text.py reads text content from arbitrary PDF files provided as input.\n
Boundary markers: No specific delimiters or instructions to ignore embedded commands are included in the prompt templates to distinguish between extracted text and instructions.\n
Capability inventory: The skill includes scripts for file reading and writing (scripts/extract_text.py) and instructions for interacting with a memory management system.\n
Sanitization: The extracted text is passed to the agent without sanitization, filtering, or escaping, allowing instructions within the PDF to potentially influence agent behavior.\n- [EXTERNAL_DOWNLOADS]: The skill documentation and reference files recommend the installation of standard, well-known PDF processing libraries (pdfplumber, pymupdf, pdfminer.six) from official package registries.\n- [COMMAND_EXECUTION]: The skill provides command-line examples for extracting text and managing memory, which involves executing Python scripts and using standard shell utilities like grep.

pdf-reader