The Agent Skills Directory

[Indirect Prompt Injection] (HIGH): The skill's primary purpose is to ingest and process external content from PDF files, creating a significant attack surface for indirect prompt injection.
Ingestion points: Untrusted data enters the agent context via pypdf.PdfReader.extract_text(), pdfplumber.open().pages.extract_text(), and OCR via pytesseract.image_to_string() (SKILL.md).
Boundary markers: None. There are no delimiters or instructions provided to the agent to treat extracted text as untrusted data.
Capability inventory: The skill possesses significant side-effect capabilities, including writing to the filesystem (PdfWriter.write, combined_df.to_excel, canvas.save) and executing shell commands via the documented CLI tools.
Sanitization: No sanitization or filtering of extracted content is implemented in the provided examples.
[External Downloads] (MEDIUM): The skill depends on numerous third-party Python packages and system utilities that are not part of a standard environment.
Evidence: Requires pypdf, pdfplumber, reportlab, pandas, pytesseract, pdf2image, and system tools like poppler-utils, qpdf, and pdftk.
[Command Execution] (MEDIUM): The skill documents and encourages the use of various command-line utilities to manipulate files on the host system.
Evidence: Shell snippets for pdftotext, qpdf, pdftk, and pdfimages are provided for file modification and extraction tasks.

pdf