← 返回 Skills 市场
146
总下载
1
收藏
0
当前安装
1
版本数
在 OpenClaw 中安装
/install socneo-autonomous-agent
功能描述
AI Autonomous Agent Framework with self-driven capabilities. Implements perception, judgment, execution, and reflection layers for intelligent autonomous ope...
安全使用建议
This skill implements a full autonomous agent that can watch file paths, monitor other skills, collect user interaction and system metrics, and store memories. Before installing or enabling autonomous operation: 1) Review the remaining/truncated source files (especially package_skill.py, any storage backend implementations, and any networking code) for outbound network calls, hardcoded endpoints, or export logic. 2) Check default config (autonomous_config.yaml) to see which file paths and event sources are enabled by default; restrict watch paths to non-sensitive directories. 3) Run the agent first in MANUAL/ASSISTED mode (not fully autonomous) and in a sandboxed environment or VM. 4) Disable or audit persistent backends — prefer in-memory operation until you confirm storage is local and encrypted (or absent). 5) Look for any code that invokes requests/sockets/ssh/ftp or packages and transmits data externally. 6) Do not grant elevated OS permissions or admin-level access until you are satisfied with the code. If you want, provide the remaining truncated files (or package_skill.py and storage-related files) and I can re-evaluate those specific parts — that would raise confidence above the current medium level.
功能分析
Type: OpenClaw Skill
Name: socneo-autonomous-agent
Version: 1.0.0
The skill bundle implements a sophisticated 'Autonomous Agent Framework' featuring a four-layer cognitive architecture: Perception, Judgment, Execution, and Reflection. The code provides comprehensive modules for system monitoring (using psutil), task decomposition, risk-based decision making, and a persistent memory system. While there are several implementation bugs where Python path expressions are incorrectly treated as string literals (e.g., in state_awareness.py, event_listener.py, and package_skill.py), the logic is entirely consistent with the stated purpose of creating a self-improving autonomous agent. No evidence of data exfiltration, malicious execution, or unauthorized persistence was found.
能力评估
Purpose & Capability
The name/description (autonomous agent) aligns with the included Python modules (perception, judgment, execution, reflection). The presence of code files implementing file watchers, skill-usage tracking, system metrics and a memory system is coherent with the stated purpose. However the SKILL.md claims instruction-only install but many code files are present — not a security flaw by itself, but it increases the runtime surface to audit.
Instruction Scope
SKILL.md and the code explicitly instruct/implement monitoring of file system, skill usage, and user activity and refer to 'event-sources all' and persistent memory. That scope allows reading and recording of arbitrary paths and user interactions. The instructions are high-level (enable event sources, start agent) which gives the agent wide discretion to collect system state unless constrained by configuration; this is scope creep compared to a narrow helper skill.
Install Mechanism
There is no install spec (lowest install risk) and no declared external downloads. That reduces supply-chain risk. Nonetheless, the package contains many executable Python modules — these will run when the skill is used, so the absence of a network install step does not eliminate runtime risk.
Credentials
The skill requests no environment variables or external credentials in registry metadata. The code uses local system APIs (os, psutil) and reads files/paths configured at runtime. There are no declared secrets requirements, which is proportionate; still, ability to monitor arbitrary files is sensitive even without explicit credentials.
Persistence & Privilege
The skill is not marked always:true (good), but it implements a memory system and references storage backend selection and retention periods. Some memory layers are 'permanent' by design. Because persistent storage/backends and any network/export logic are in the omitted/truncated files, there's risk the skill could write sensitive data to disk or external services. Treat persistent memory and any packaging/telemetry code (e.g., package_skill.py) as higher-privilege features that must be audited.
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 socneo-autonomous-agent - 安装完成后，直接呼叫该 Skill 的名称或使用
/socneo-autonomous-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
- Initial release of the Autonomous Agent framework (v1.0.0).
- Implements a four-layer cognitive architecture: Perception, Judgment, Execution, and Reflection.
- Features adaptive monitoring, risk-aware autonomous decision-making, resilient task execution, and continuous learning.
- Includes robust safety mechanisms: permission management, audit logging, user override, and explainable decisions.
- Supports advanced configuration, external integrations, and gradual enablement for safe deployment.
- Comprehensive documentation with usage scenarios, troubleshooting, and best practices.
元数据
常见问题
Autonomous Agent 是什么？
AI Autonomous Agent Framework with self-driven capabilities. Implements perception, judgment, execution, and reflection layers for intelligent autonomous ope... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 146 次。
如何安装 Autonomous Agent？
在 OpenClaw 或 Claude Code 对话框中运行命令「/install socneo-autonomous-agent」即可一键安装，无需额外配置。
Autonomous Agent 是免费的吗？
是的，Autonomous Agent 完全免费，采用 MIT-0 许可证，可自由下载、安装和使用。
Autonomous Agent 支持哪些平台？
Autonomous 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 Autonomous Agent？
由 socneo（@socneo）开发并维护，当前版本 v1.0.0。
推荐 Skills