kimi-agent-policy
Kimi (Moonshot AI) agent tool-use policy ported to OpenClaw. Covers step limits, web search, image search, data sources, ipython, memory, content display, and capability boundaries. All Kimi-specific tools are mapped to available OpenClaw skills.
Install via CLI (Recommended)
clawhub install openclaw/skills/skills/2233admin/kimi-agent-policyWhat This Skill Does
kimi-agent-policy serves as a standardized operational framework for OpenClaw agents, mirroring the sophisticated tool-use discipline of the Moonshot AI Kimi ecosystem. It enforces strict guidelines on how an agent interacts with external tools, optimizing for reliability, efficiency, and resource management. This skill maps complex Kimi-specific capabilities into native OpenClaw tools—specifically leveraging the agent-browser, built-in search, and internal memory utilities—to ensure that agent behavior remains predictable and within safety constraints while maintaining high performance without requiring third-party API keys.
Installation
To integrate this policy into your agent, use the OpenClaw management CLI:
clawhub install openclaw/skills/skills/2233admin/kimi-agent-policy
Use Cases
- Structured Research: Implementing a standardized workflow where an agent performs a search, followed by a targeted URL fetch, and a final synthesis.
- Constraint Management: Automatically restricting agents to a maximum of 10 tool calls per turn and limiting web searches to prevent loop errors or excessive token usage.
- Privacy-First Browsing: Utilizing the
agent-browserandddgrtools to conduct information gathering without exposing data to external API providers. - System Governance: Ensuring that all agents within a fleet follow the same rules regarding image processing, data sourcing, and memory handling.
Example Prompts
- "Look up the latest updates on the OpenClaw release notes and summarize the new features for me."
- "Search for images of the Eiffel Tower and provide a brief historical overview of the architecture."
- "Follow thekimi-agent-policy for our next analysis project; keep step counts within the 10-step limit."
Tips & Limitations
- Step Limits: Always remember that the 10-step limit is per turn. If a complex task needs more, instruct the agent to provide a partial status update and continue in a follow-up turn.
- Browser Efficiency: While
agent-browseris powerful, always refine your search queries first usingweb_searchorddgrto avoid fetching unnecessary pages. - No API Keys: This policy focuses on open-source and free-to-use tools. If you require advanced features from paid APIs, you will need to augment this policy with additional authentication-based skills.
Metadata
Not sure this is the right skill?
Describe what you want to build — we'll match you to the best skill from 16,000+ options.
Find the right skillPaste this into your clawhub.json to enable this plugin.
{
"plugins": {
"official-2233admin-kimi-agent-policy": {
"enabled": true,
"auto_update": true
}
}
}Tags(AI)
Flags: network-access, file-read, code-execution
Related Skills
Api Monitor
Skill by 2233admin
markdown-browser
Wrapper skill for OpenClaw web_fetch results. Use when you need MECE post-processing on fetched pages: policy decision from Content-Signal, privacy redaction, optional markdown normalization fallback, and stable output schema without re-implementing network fetch.
Clawapi Manager
Skill by 2233admin
evomap-node-controller
管理 EvoMap 节点的启动、配置和监控。用于在服务器上启动/停止 EvoMap evolver 循环、配置节点 ID，处理节点绑定等。使用场景：用户要求启动/停止 EvoMap 节点、查看节点状态、配置节点 ID、解决节点连接问题。⚠️ 使用前必须配置环境变量。
adaptive-review
Adaptive code review that routes to haiku/sonnet/opus based on diff complexity signals. Use instead of requesting-code-review for cost-efficient reviews.