Pdfreader

Extract text and metadata from PDF files using PyMuPDF, supporting large files and outputting results in JSON format.

This skill appears to do what it claims: extract text and metadata from PDFs using PyMuPDF. Before installing or running it, consider: 1) Run pip install pymupdf in an isolated environment (virtualenv/container) — PyMuPDF includes compiled code from PyPI. 2) The script enforces 'within current working directory' but allows subdirectories and does not resolve symlinks; avoid placing untrusted symlinks inside the working directory to prevent escapes. 3) Because the source/homepage is unknown, prefer running the script in a sandbox and review the code yourself (or run it on non-sensitive PDFs) before giving it access to important files. If you need stricter confinement (no subdirectories or symlink protections), request a code change to use os.path.realpath checks and a configurable safe directory.

Type: OpenClaw Skill Name: pdfreader Version: 1.0.3 The OpenClaw skill bundle is designed to extract text from PDF files using PyMuPDF. The `SKILL.md` documentation provides clear, non-malicious instructions and explicitly states security restrictions. The `pdf_reader.py` script implements robust path validation (`is_safe_input_path`, `is_safe_output_path`) to prevent path traversal and restrict file operations to the current working directory and specific file types (.pdf for input, .json for output). There are no signs of data exfiltration, malicious execution, persistence, or prompt injection attempts against the agent. The code is well-contained and aligns with its stated purpose and security measures.