Back to skill
Skillv1.1.2
ClawScan security
Gemini Watermark Remover · ClawHub's context-aware review of the artifact, metadata, and declared behavior.
Scanner verdict
BenignMar 6, 2026, 7:18 PM
- Verdict
- benign
- Confidence
- high
- Model
- gpt-5-mini
- Summary
- The package is internally consistent: the description, runtime instructions, and included Python script all align with a local, offline Gemini watermark removal tool and do not request unrelated credentials or installs.
- Guidance
- This skill appears coherent and local-only, but before running any third-party script: (1) inspect the full scripts/remove_watermark.py for any unexpected network calls or filesystem operations (the provided excerpt looks local-only), (2) run in an isolated environment or VM if you are unsure, (3) confirm you have the legal right to remove watermarks for the images you process, and (4) install dependencies into a virtualenv to avoid affecting system Python packages.
- Findings
[base64-block] expected: SKILL.md and the script embed watermark templates as Base64 PNGs; the scanner's base64-block hit is expected for embedded image assets and not necessarily malicious.
Review Dimensions
- Purpose & Capability
- okName/description claim removing Gemini visible watermarks; included script and README implement a reverse alpha-blending algorithm with embedded watermark templates. No unrelated binaries, env vars, or external services are requested.
- Instruction Scope
- okSKILL.md instructs only local usage (pip install Pillow/numpy and run scripts/remove_watermark.py). It does not direct reading unrelated system files, sending data to remote endpoints, or accessing credentials. The script embeds templates and operates on local images.
- Install Mechanism
- okNo install spec; this is instruction-only with a bundled Python script and requirements.txt. Dependency installation via pip is reasonable and expected for a Python image-processing tool. No remote downloads or nonstandard installers are used.
- Credentials
- okNo environment variables, credentials, or config paths are required. The requested dependencies (Pillow, NumPy) are appropriate for image processing.
- Persistence & Privilege
- okSkill does not request always:true or modify system/other-skill configs. Default invocation settings are used; the skill is user-invocable and not forced into all agent runs.