pdf

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.80). The guide embeds plaintext passwords in examples/commands (e.g., qpdf --password=mypassword and writer.encrypt("userpassword","ownerpassword")), which encourages including secret values verbatim in generated code or commands.

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (medium risk: 0.65). The skill’s runtime workflow ingests outsider-authored PDF content (e.g., the user-supplied <input.pdf> body text/graphics) via scripts like scripts/check_fillable_fields.py (calls PdfReader.get_fields() on the PDF) and scripts/extract_form_field_info.py / scripts/extract_form_structure.py (uses pdfplumber to extract words/lines/rects), which can include arbitrary free text from the PDF into the LLM context indirectly through extracted JSON/text artifacts.