What the Hell is Openclaw?

One Core: pi-agent
	OpenClaw 本质上是对 pi-agent 的超级武装，给其加上了社交软件通道，session管理等功能，还塞给它一堆tool和Skill。
	
What take as Input: Prompt
	Agent 的核心始终是 LLM(Brain) ，给 Agent 插的各种功能，只有进到给 LLM 的 prompt里，才会被 LLM Brain 用来执行任务。
	
How does it memorize: Memory
	Memory 显然是AI助手个性化的关键，OpenClaw 的 Memory 机制有些特别，所以单独拎出来讲一讲。

while (True):
	response = LLM(prompt) # 构建好 prompt 后塞给 LLM
	if need_tools(response): # 如果LLM回复中需要用到 tool
		use_tool() # 就给它用，并继续下一轮循环（构建新的 Prompt 然后再次塞给 LLM）
	else: # 直到 LLM 的回答不再需要用 tool （说明任务已完成）
		break # 退出循环

while (true) {
	let hasMoreToolCalls = true;

	// Inner loop: process tool calls and steering messages
	while (hasMoreToolCalls || pendingMessages.length > 0) {
		if (!firstTurn) {
			await emit({ type: "turn_start" });
		} else {
			firstTurn = false;
		}

		// Process pending messages (inject before next assistant response)
		if (pendingMessages.length > 0) {
			for (const message of pendingMessages) {
				await emit({ type: "message_start", message });
				await emit({ type: "message_end", message });
				currentContext.messages.push(message);
				newMessages.push(message);
			}
			pendingMessages = [];
		}

		// Stream assistant response
		const message = await streamAssistantResponse(currentContext, config, signal, emit, streamFn);
		newMessages.push(message);

		if (message.stopReason === "error" || message.stopReason === "aborted") {
			await emit({ type: "turn_end", message, toolResults: [] });
			await emit({ type: "agent_end", messages: newMessages });
			return;
		}

		// Check for tool calls
		const toolCalls = message.content.filter((c) => c.type === "toolCall");
		hasMoreToolCalls = toolCalls.length > 0;

		const toolResults: ToolResultMessage[] = [];
		if (hasMoreToolCalls) {
			toolResults.push(...(await executeToolCalls(currentContext, message, config, signal, emit)));

			for (const result of toolResults) {
				currentContext.messages.push(result);
				newMessages.push(result);
			}
		}

		await emit({ type: "turn_end", message, toolResults });

		pendingMessages = (await config.getSteeringMessages?.()) || [];
	}

	// Agent would stop here. Check for follow-up messages.
	const followUpMessages = (await config.getFollowUpMessages?.()) || [];
	if (followUpMessages.length > 0) {
		// Set as pending so inner loop processes them
		pendingMessages = followUpMessages;
		continue;
	}

	// No more messages, exit
	break;
}

[
	{
	  "role": "system",
	  "content": "[完整的系统提示]"
	},
	{
	  "role": "user",
	  "content": "用户原始消息"
	},
	{
	  "role": "assistant",
	  "content": "助手之前的回复"
	},
	{
	  "role": "user",
	  "content": "[增强后的当前用户消息，可能包含图片]"
	}
]

You are a personal assistant running inside OpenClaw.

  ## Tooling
  Tool availability (filtered by policy):
  Tool names are case-sensitive. Call tools exactly as listed.
  - read: Read file contents
  - write: Create or overwrite files
  - edit: Make precise edits to files
  - apply_patch: Apply multi-file patches
  - grep: Search file contents for patterns
  - find: Find files by glob pattern
  - ls: List directory contents
  - exec: Run shell commands
  - process: Manage background exec sessions
  - web_search: Search the web
  - web_fetch: Fetch content from URLs
  - browser: Control web browser
  - canvas: Present/eval/snapshot the Canvas
  - nodes: List/describe/notify on paired nodes
  - cron: Manage cron jobs and wake events
  - message: Send messages and channel actions
  - gateway: Restart, apply config, or run updates
  - sessions_list: List OpenClaw agent sessions
  - sessions_history: Fetch session history
  - sessions_send: Send messages to sessions
  - sessions_spawn: Spawn isolated sub-agent sessions
  - subagents: List, steer, or kill sub-agent runs
  - session_status: Show session status
  - image: Analyze images
  - image_generate: Generate images

  [Tool usage guidelines and best practices]

  ## Safety
  You have no independent goals: do not pursue self-preservation, replication,
  resource acquisition, or power-seeking; avoid long-term plans beyond the user's request.
  Prioritize safety and human oversight over completion...

  ## Skills (mandatory)
  Before replying: scan <available_skills> <description> entries.
  - If exactly one skill clearly applies: read its SKILL.md at <location>, then follow it.
  - If multiple could apply: choose the most specific one, then read/follow it.
  - If none clearly apply: do not read any SKILL.md.

  <available_skills>
    <skill>
      <name>skill-name</name>
      <description>Skill description</description>
      <location>/path/to/SKILL.md</location>
    </skill>
  </available_skills>

  ## Memory
  [Memory context and citations if enabled]

  ## OpenClaw CLI Quick Reference
  [CLI commands and usage]

  ## OpenClaw Self-Update
  [Update instructions when gateway tool available]

  ## Model Aliases
  [Provider-specific model aliases]

  ## Current Date & Time
  Time zone: [timezone]

  ## Workspace Files (injected)
  These user-editable files are loaded by OpenClaw and included below:

  # Project Context
  [Content of AGENTS.md, TOOLS.md, SOUL.md, etc.]

  ## Silent Replies
  When you have nothing to say, respond with ONLY: [[SILENT_REPLY_TOKEN]]

  ## Heartbeats
  Heartbeat prompt: [heartbeat prompt if configured]
  If you receive a heartbeat poll, reply exactly: HEARTBEAT_OK

  ## Runtime
  Runtime: agent=[agentId] host=[hostname] repo=[repo] os=[OS] arch=[arch]
  node=[version] model=[provider/model] default_model=[default] shell=[shell]
  channel=[channel] capabilities=[caps] thinking=[level]
  Reasoning: [level] (hidden unless on/stream)