← 返回 Skills 市场
My Proactive Agent
作者
liujiang817
· GitHub ↗
· v1.0.0
· MIT-0
230
总下载
0
收藏
0
当前安装
1
版本数
在 OpenClaw 中安装
/install my-proactive-agent
功能描述
Transform AI agents from task-followers into proactive partners that anticipate needs and continuously improve. Now with WAL Protocol, Working Buffer, Autono...
安全使用建议
This skill looks like a coherent 'proactive agent' architecture (it writes/reads onboarding, memory, heartbeat files and includes a local security-audit script). But before installing: 1) Review scripts/security-audit.sh and other files yourself — they read $HOME config and /tmp logs; understand what will be read. 2) Run the skill in an isolated or sandboxed workspace (not your full home) first. 3) Confirm .credentials is actually gitignored and that credentials remain out of tracked files. 4) Clarify how 'proactive' actions are gated — ensure anything that leaves the machine (emails, network posts, external agents) requires explicit human approval. 5) Because the repo contains many prompt-injection examples (intended for detection), verify the agent runtime won't confuse them for system prompts. If you lack isolation, don't install this skill on a machine with sensitive data until you've audited it.
功能分析
Type: OpenClaw Skill
Name: my-proactive-agent
Version: 1.0.0
The 'proactive-agent' skill bundle is a comprehensive framework designed to enhance AI agent autonomy, persistence, and self-improvement through structured protocols like WAL (Write-Ahead Logging) and Working Buffers. It includes robust security instructions in SKILL.md and AGENTS.md that specifically warn against prompt injection from external content and data exfiltration. The bundle also provides a legitimate security utility, scripts/security-audit.sh, which performs local checks for exposed secrets and improper file permissions without any evidence of malicious intent or unauthorized data transmission.
能力评估
Purpose & Capability
Name/description (proactive, WAL, working buffer, heartbeats) align with the shipped assets (ONBOARDING.md, AGENTS.md, MEMORY.md, heartbeat guidance) and the included security-audit script. Nothing required (no env vars, no binaries) is obviously unrelated to the stated goal.
Instruction Scope
SKILL.md and the assets instruct the agent to read/write many workspace files (ONBOARDING.md, USER.md, SESSION-STATE.md, memory/*) which is consistent with the purpose, but they also reference and suggest inspecting files outside the workspace (e.g., tail /tmp/clawdbot/*.log in HEARTBEAT.md, $HOME/.clawdbot/clawdbot.json in security-audit.sh). The skill includes detection lists for prompt-injection phrases; those phrases appear in the files (as detection rules) but the presence of many instruction-like strings increases risk if the agent mis-parses context. There are contradictory guardrails in different files (e.g., 'Don't ask permission. Just do it.' vs 'Nothing external without approval'), which could result in unintended autonomous actions depending on how the agent implements them.
Install Mechanism
No install spec; this is instruction-only with a single audit script. No downloads or archive extraction. The included shell script is readable and does local checks — no external code fetches or network installs present.
Credentials
The skill declares no required credentials or env vars (appropriate), but its runtime guidance and audit script reference user home config ($HOME/.clawdbot/clawdbot.json), /tmp logs, and a .credentials directory. Those references mean the skill expects read/write access beyond just the working directory, which is reasonable for a proactive agent but should be considered a higher privilege than a purely in-workspace helper.
Persistence & Privilege
Metadata does not request 'always:true' and autonomous invocation is allowed (platform default). The content encourages proactive, persistent behavior (writing many files, heartbeats, self-healing). Because the skill advocates some 'do it' behaviors, ensure external actions remain gated by human approval as other files recommend.
如何使用
- 确保已安装 OpenClaw（本地或 Docker 部署）
- 在对话框中输入安装命令：
/install my-proactive-agent - 安装完成后，直接呼叫该 Skill 的名称或使用
/my-proactive-agent触发 - 根据 Skill 的参数说明提供必要输入，即可获得结构化输出
版本历史
v1.0.0
proactive-agent v3.1.0 introduces enhanced autonomy and reliability features for AI agents.
- Added support for autonomous vs prompted crons with guidance on using `systemEvent` and `isolated agentTurn`.
- New protocols: WAL Protocol (Write-Ahead Logging), Working Buffer, and Compaction Recovery ensure critical details survive context loss.
- Security hardening expanded, including tool migration checklist and strengthened context management.
- Emphasizes verifying implementation over intent to improve reliability.
- Documentation updated with new patterns and self-improvement guardrails.
元数据
常见问题
My Proactive Agent 是什么？
Transform AI agents from task-followers into proactive partners that anticipate needs and continuously improve. Now with WAL Protocol, Working Buffer, Autono... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 230 次。
如何安装 My Proactive Agent？
在 OpenClaw 或 Claude Code 对话框中运行命令「/install my-proactive-agent」即可一键安装，无需额外配置。
My Proactive Agent 是免费的吗？
是的，My Proactive Agent 完全免费，采用 MIT-0 许可证，可自由下载、安装和使用。
My Proactive Agent 支持哪些平台？
My Proactive Agent 跨平台运行，可在任意部署了 OpenClaw / Claude Code 的环境中使用（cross-platform）。
谁开发了 My Proactive Agent？
由 liujiang817（@liujiang817）开发并维护，当前版本 v1.0.0。
推荐 Skills