Proactive Agent

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

This skill is mostly coherent with a 'proactive agent' design, but there are a few things to check before installing or running anything: 1) Inspect ~/.openclaw/.env and confirm it doesn't contain secrets you wouldn't want the skill to access; the TOOLS.md suggests keys may live there. 2) Review scripts/security-audit.sh and any other executable files locally (don't run them until you review). 3) Decide whether you are comfortable letting an agent read and write workspace files (SESSION-STATE.md, MEMORY.md, USER.md, etc.). 4) Clarify the contradictory guidance in the docs (some files say 'Don't ask permission. Just do it.' while others require approval before external actions). 5) If you will enable external APIs (Tavily or browser control), add only the minimal credentials and consider running the skill in an isolated workspace or container first. 6) Prefer getting a known source/repo or author verification (the registry lists an opaque owner) — if you cannot verify the origin, treat it as higher risk. If you want, I can (a) produce a short checklist of exact lines to look for in the script(s) that would indicate network exfiltration, or (b) walk you through running the security-audit.sh safely in a disposable environment.

Type: OpenClaw Skill Name: proactive-agent-cp3d Version: 1.0.0 The 'proactive-agent' skill bundle is a comprehensive framework designed to enhance AI agent autonomy, state management, and reliability. It introduces structured protocols like WAL (Write-Ahead Logging) and Working Buffer to handle context limits, alongside a defensive security audit script (scripts/security-audit.sh) that scans for exposed credentials and risky code patterns. The documentation (SKILL.md, AGENTS.md, and security-patterns.md) explicitly instructs the agent to defend against prompt injection, avoid unauthorized data exfiltration, and verify actions before reporting completion, showing a strong alignment with security best practices rather than malicious intent.