Back to skill
Skillv2026.3.26
ClawScan security
Input Refiner · ClawHub's context-aware review of the artifact, metadata, and declared behavior.
Scanner verdict
BenignMay 1, 2026, 8:23 AM
- Verdict
- Benign
- Confidence
- high
- Model
- gpt-5.5
- Summary
- This instruction-only skill is focused on clarifying user input and shows no evidence of code execution, credential use, persistence, or external data access.
- Guidance
- This skill appears safe to install for rewriting and organizing user input. As with any rewriting tool, users should review the refined output to ensure it still matches their intent before using it as an instruction for another agent or model.
Review Dimensions
- Purpose & Capability
- okThe stated purpose and instructions are coherent: the skill rewrites vague or unstructured input into clearer form while repeatedly requiring preservation of the user's original intent.
- Instruction Scope
- okThe instructions include clear boundaries against inventing requirements, changing goals, or over-expanding simple requests.
- Install Mechanism
- okThere is no install specification, no code file, and no required binaries or environment variables.
- Credentials
- okThe skill does not request filesystem, network, shell, credential, account, or device access.
- Persistence & Privilege
- okThe artifacts show no persistence mechanism, background behavior, elevated privilege, memory storage, or autonomous mutation authority.