Back to skill
Skillv3.4.1
ClawScan security
Prototype · ClawHub's context-aware review of the artifact, metadata, and declared behavior.
Scanner verdict
BenignMar 17, 2026, 7:14 PM
- Verdict
- Benign
- Confidence
- high
- Model
- gpt-5-mini
- Summary
- The skill's declared purpose (building HTML prototypes) matches its code and instructions: it is a bash-based generator that writes/updates HTML and CSS files and does not request credentials, installs, or network access.
- Guidance
- This skill appears coherent and implements what it advertises: a bash-based HTML prototype generator. Before running or allowing autonomous execution, review scripts/script.sh yourself and avoid pointing the tool at system or sensitive directories (e.g., /etc, your home with secrets) because it will create and overwrite files in the output/input paths you supply. If you want extra safety, run the commands in a sandboxed workspace or container and do not run them with elevated privileges.
Review Dimensions
- Purpose & Capability
- okName/description (interactive HTML prototypes) aligns with the included script and commands (create, component, animate, link, preview, export). The only declared requirement is bash 4+, which is appropriate for a shell script-based tool.
- Instruction Scope
- okSKILL.md tells the agent to run scripts/script.sh with explicit commands and options. The script generates and updates HTML/CSS/JS files and prints previews; it does not request reading unrelated system files or contacting external endpoints. Note: the script writes files to user-specified output paths and reads prototype input directories (for preview/export), so pointing it at sensitive system paths could overwrite or expose files — this is expected behavior for a file generator but worth caution.
- Install Mechanism
- okNo install specification is provided (instruction-only with a bundled script). Nothing is downloaded or written to disk beyond the included script when the skill is installed; execution happens when the agent runs the script. This is low-risk compared to remote installs.
- Credentials
- okThe skill does not request environment variables, credentials, or config paths. The script does not appear to read secrets or external tokens, which fits the stated purpose.
- Persistence & Privilege
- okalways: false and no special persistence or modification of other skills or agent configuration are requested. The skill can be invoked by the model (normal default) but does not demand elevated or permanent privileges.