claw-prompt-injection-guard

• Risk level: Low
• Analyzed: May 1, 2026, 6:48 AM
• Findings: 0
• Version: 1.0.0

Overview

This is an instruction-only safety skill that teaches the agent to treat external content cautiously and ask before sensitive actions, with no code, credentials, install steps, or persistence shown.

From the provided artifacts, this skill appears safe and purpose-aligned. Users should expect it to make the agent more cautious around webpages, emails, search results, and other external content, including asking for confirmation before sensitive actions.