Proactive Agent
Transform AI agents from task-followers into proactive partners that anticipate needs and continuously improve. Now with WAL Protocol, Working Buffer, Autonomous Crons, and battle-tested patterns. Part of the Hal Stack 🦞
v3.1.0
2026/02/05
Added: Autonomous vs Prompted Crons, Verify Implementation Not Intent, Tool Migration Checklist
Security Scan
Status
suspicious
OpenClaw
gpt-5.5
OpenClaw analysis
The skill is not clearly malicious, but it gives a proactive agent broad persistent-memory, background-work, local-cleanup, and email/calendar-checking authority without enough user control.
Confidence: high
VirusTotal
Type: OpenClaw Skill Name: proactive-agent Version: 3.1.0 This skill bundle is designed to create a proactive, self-improving AI agent with extensive security hardening. Multiple files (SKILL.md, AGENTS.md, SOUL.md, references/security-patterns.md) contain explicit instructions for the agent to defend against prompt injection, prevent data exfiltration, confirm deletions, and vet other skills for malicious content. The included `scripts/security-audit.sh` actively scans for exposed secrets, insecure permissions, and checks for security best practices. While the agent is instructed to use broad capabilities like CLI access, web search, and cron jobs, these are consistently framed within the context of problem-solving for the human, with strong guardrails against unauthorized external actions. There is no evidence of intentional harmful behavior; rather, the content demonstrates a strong focus on security and safe operation.