Back to skill
Skillv1.0.0
ClawScan security
Agent Arch · ClawHub's context-aware review of the artifact, metadata, and declared behavior.
Scanner verdict
BenignMay 1, 2026, 8:28 AM
- Verdict
- Benign
- Confidence
- high
- Model
- gpt-5.5
- Summary
- This is an instruction-only guidance skill for designing AI agents, with no code, credentials, install steps, persistence, or privileged access.
- Guidance
- This skill appears safe to install as a guidance-only workflow. It should simply help structure conversations about agent architecture, safety boundaries, validation, and operations.
Review Dimensions
- Purpose & Capability
- okThe stated purpose and SKILL.md content align: it provides a structured workflow for agent architecture discussions.
- Instruction Scope
- okThe workflow asks the assistant to offer stages, ask clarifying questions, and adapt if the user declines; it does not override user intent or force unsafe tool use.
- Install Mechanism
- okThere is no install specification and no code files; the skill is instruction-only.
- Credentials
- okNo binaries, environment variables, credentials, config paths, network access, or OS-specific capabilities are requested.
- Persistence & Privilege
- okThe artifacts show no persistence, background execution, account access, memory storage, or privilege escalation.