The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes untrusted PDF files, creating an attack surface for indirect prompt injection where malicious text in a document could manipulate agent behavior.
Ingestion points: SKILL.md uses pdfplumber.open() in functions like parse_resume and extract_invoice_data to read content from user-provided files.
Boundary markers: The instructions do not define delimiters or provide specific prompts to the agent to ignore instructions embedded within the extracted text.
Capability inventory: The skill has access to code_execution and file_operations tools (identified in the frontmatter), which increases the impact of a potential injection attack.
Sanitization: The provided code templates do not perform sanitization or filtering of the text extracted from PDFs before it is used by the agent for further processing or decision-making.

pdf-extraction