The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill instructions require the agent to generate Python code for PDF creation using the ReportLab library, write this code to a local file (e.g., generate_pdf.py), and execute it. To mitigate potential rendering issues, the skill mandates running a provided sanitization script (scripts/sanitize_code.py) on the generated file before execution to remove forbidden Unicode characters and escapes.
[INDIRECT_PROMPT_INJECTION]: The skill provides numerous ingestion points for untrusted data by reading, OCR-ing, and extracting metadata from user-provided PDF files via pypdf, pdfplumber, and pytesseract.
Ingestion points: Files processed through scripts/extract_form_field_info.py, scripts/fill_fillable_fields.py, and scripts/convert_pdf_to_images.py.
Boundary markers: The instructions do not define strict delimiters for data extracted from PDFs, although they do instruct the agent to verify factual content via web searches.
Capability inventory: The agent can perform file system writes and execute several command-line tools such as qpdf and pdftotext.
Sanitization: Code sanitization is performed on generated scripts, but no specific sanitization of text content extracted from input PDFs is mentioned.
[COMMAND_EXECUTION]: The skill documentation and helper scripts utilize several external CLI utilities for PDF processing, including pdftotext, qpdf, pdftoppm, and pdfimages. These tools are invoked via subprocess calls or direct instructions to perform page manipulation and text extraction.

pdf