PDF

PDF workflow

This is workflow guidance, not a callable tool. Use the shell or code-execution tool exposed by the active OIX harness. Do not look for a tool named pdf.

Workflow

1. Inspect metadata and text with pypdf or pdfplumber.
2. Render pages to PNG with Poppler (pdftoppm) before making layout-sensitive decisions.
3. Use pypdf for page operations and forms, reportlab for new PDFs, and pdfplumber for focused extraction.
4. After every meaningful write, reopen the file and render all affected pages.
5. If the file is open in Interpreter, call interpreter-app tools builtin-interpreter interpreter_refresh_file --json '{"path":"/absolute/path.pdf"}' once after the final write.

Dependencies

uv pip install pypdf pdfplumber reportlab

Use an existing Python and Poppler installation when available. Install only missing pieces, and request approval for exact system-level installation commands.

Quality bar

• No clipped, overlapping, missing, or unreadable content.
• Preserve page size, rotation, form field identity, and source fidelity.
• Verify page count and expected text after saving.
• Inspect the latest rendered pages before delivery.
• Keep intermediate renders and scripts out of the final deliverable folder.

Derived from OpenAI's Apache-2.0 pdf skill and modified for OIX code execution and Workstation refresh behavior.