The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external PDF and LaTeX files which serves as a vulnerability surface for indirect prompt injection attacks.
Ingestion points: extract_pdf_text.py and parse_pdf_sections.py ingest content from user-provided paper files.
Boundary markers: The workflow does not explicitly use delimiters or instructions to ignore potential injections embedded within the extracted text.
Capability inventory: The skill executes local Python scripts and writes output files using shell commands in SKILL.md.
Sanitization: No evidence of sanitization or filtering of the extracted text for prompt injection patterns was found in the provided scripts.

self-review