Watermark Pro
Analysis
This looks like a local watermarking helper with no credential or network behavior, but it edits user-selected files and relies on unpinned Python packages.
Findings (4)
Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.
Checks for instructions or behavior that redirect the agent, misuse tools, execute unexpected code, cascade across systems, exploit user trust, or continue outside the intended task.
for para in header.paragraphs: para.clear() ... doc.save(output_path)
The Word watermark code clears existing header paragraphs before saving the output, which is a local file mutation users may not expect when simply adding a watermark.
temp_img = os.path.join(os.path.dirname(output_path), '_temp_wm.png') ... wm_img.save(temp_img) ... os.remove(temp_img)
The PDF watermark path writes a fixed-name temporary PNG in the output directory; this is scoped and local, but could overwrite an existing file with that name or leave a temp file if processing fails.
dependencies: "pip install pillow python-docx python-pptx pymupdf"
The skill depends on third-party Python packages without pinned versions; these libraries are expected for the stated purpose, but dependency provenance and reproducibility are not locked.
- 纯本地处理,无隐私风险
The artifacts support local processing and show no network calls, but the phrase 'no privacy risk' is an absolute assurance; local documents can still contain sensitive information and outputs need careful handling.
