photo-watermark

为照片添加可自定义的文字或图片水印。在处理图片、添加水印或批量处理照片时使用。支持JPG、PNG和其他常见图片格式。

Overall the package appears to implement what it claims: local watermarking and optional retouching, using standard Python imaging libraries with no network calls or credential access. Before installing or running it: 1) Inspect scripts/add_watermark.py and the is_safe_path() implementation to confirm path validation blocks undesired files (the tests reference this). 2) Note documentation/metadata inconsistencies: registry metadata lists version 1.0.5 while many docs/manifests reference v2.1.0/2.2.0 and mention an install.py that is not present — confirm you have the intended release. 3) Verify any optional dependency (opencv/numpy/scipy) is installed only if you need retouching; opencv can be large. 4) If you plan to run any curl|sh installation examples, download and review the script first (as the README suggests). 5) Run the included test_skill.py in a safe environment to validate behavior on your machine before processing important files.

Type: OpenClaw Skill Name: photo-watermark Version: 1.0.5 The photo-watermark skill bundle is a legitimate image processing utility for adding watermarks and performing skin retouching. The code demonstrates high security awareness, featuring a robust path validation function (is_safe_path in scripts/add_watermark.py) that prevents path traversal and restricts access to sensitive system directories. It utilizes standard libraries like Pillow, OpenCV, and NumPy for its stated functions, and the documentation (README.md, SECURITY_FIXES.md) explicitly warns users about security best practices and documents the removal of potentially risky subprocess calls. No evidence of data exfiltration, malicious execution, or prompt injection was found.