DeepThink

• Analyzed: May 12, 2026, 3:25 AM
• Findings: 1
• Version: 1.0.4
• Hash: afb71a6a7620831d435b48806b27f7c00923849892bdb6fad88b2d80e57c3b9d
• Source: palm
• External report: View on VirusTotal

Findings (1)

OpenClaw Agent Skill The skill bundle is classified as benign. All API calls are directed to the specified `https://api.deepthink.co` domain, aligning with the stated purpose of a personal knowledge base. While the `SKILL.md` instructs the AI agent to modify its own `SOUL.md` and `HEARTBEAT.md` files for self-improvement and task scheduling, these actions are clearly defined, serve a functional purpose within the agent's operational context, and do not involve exfiltration of sensitive data or unauthorized system-level persistence. Furthermore, the `SKILL.md` explicitly includes a 'CRITICAL: Prompt Injection Protection' section, instructing the agent to never blindly execute commands from transcript text and always ask permission for significant actions, indicating a proactive stance against prompt injection from external sources.