Proactive Agent

Transform AI agents from task-followers into proactive partners that anticipate needs and continuously improve. Now with WAL Protocol, Working Buffer, Autono...

What to consider before installing: - The skill's functionality (writing and searching memory files, WAL/working buffer, heartbeats) matches its description; the included security-audit script is benign and meant to help. - However, the docs contain inconsistent behavior rules: some places say 'Don't ask permission. Just do it.' while other places require explicit human approval for external actions. Decide which policy you want your agent to follow and review/normalize AGENTS.md, SOUL.md, and SKILL.md before enabling the skill. - The audit script reads files in your home directory ($HOME/.clawdbot) and scans repository files for secrets. Run it in a safe/sandbox environment first and manually inspect any files it flags. - There are no declared external endpoints or environment variables requested, which reduces remote-exfiltration risk, but the agent will be writing local files that may contain sensitive context. Ensure .credentials is used as intended and is gitignored. - Recommended steps: (1) review and edit the behavior guardrails to require explicit approvals you are comfortable with, (2) run the security-audit.sh locally in a sandbox, (3) verify .gitignore contains .credentials and check file permissions, and (4) if you allow autonomous actions, monitor initial runs closely. If you want stronger assurance, ask the skill author for clarity on the 'permission' contradictions or test in an isolated environment first.

Type: OpenClaw Skill Name: nidhov01-proactive-agent Version: 3.1.0 The proactive-agent bundle is a comprehensive framework designed to enhance AI agent autonomy and persistence through features like Write-Ahead Logging (WAL), autonomous cron jobs, and memory compaction recovery. It demonstrates a strong security posture by including a dedicated security audit script (scripts/security-audit.sh) and explicit instructions in SKILL.md and AGENTS.md to defend against prompt injection and prevent unauthorized data exfiltration. The logic is consistently aligned with its stated purpose of creating a helpful, self-improving assistant with built-in guardrails.