ClawGuard by jugaad-lab

• Analyzed: Apr 30, 2026, 3:34 AM
• Findings: 0
• Version: 1.3.1
• Hash: be78c6b392b4b417dd23e13c07dd45ebab80986ea9f051a3a4d052634466eab1
• Source: palm
• External report: View on VirusTotal

Overview

Type: OpenClaw Skill Name: jugaad-clawguard Version: 1.3.1 This skill, ClawGuard, is a security blacklist system designed to protect AI agents from malicious skills, scams, and prompt injection. The code and documentation consistently demonstrate an intent to detect and prevent harmful activities, rather than perform them. Key indicators supporting this classification include: the `lib/detector.js` module explicitly defines and checks for prompt injection patterns (e.g., 'ignore previous instructions'), dangerous commands (e.g., `curl | bash`, `rm -rf`), and known malicious URLs/domains. The `lib/mcp-scanner.js` module scans for secret exposure and command injection in *other* configurations, redacting sensitive data for reporting. The `SKILL.md` and `openclaw-plugin.js` provide instructions to the AI agent on how to *respond* to detected threats (block, warn, seek human approval via Discord), which is aligned with its security purpose. File system operations are confined to local storage (`~/.clawguard`) for configuration, audit logs, and the threat database, and network calls are for syncing the threat database from a specified GitHub repository or sending Discord approval requests, all of which are legitimate functions of a security tool.