Huo15 Openclaw Explore Mode

深度探索模式 — 系统性调研代码库、系统或话题，只读不改。借鉴 Claude Code 的 Explore Agent。

Security Scan

This is an instruction-only, read‑only ‘explore mode’ for investigating codebases; its requirements and behavior are coherent aside from a small note about using git without declaring git as a required binary.

Purpose & Capability

The name/description (deep, read-only exploration of codebases) matches the instructions: directory scanning, reading README/configs/tests, tracing call chains, and producing a structured report. Minor mismatch: the SKILL.md explicitly says to consult git log, which implies needing git, but the skill declares no required binaries.

Instruction Scope

Instructions stay within the stated purpose: they direct the agent to read files, examine git history, and produce a report. They explicitly require only read actions ('只读不改'). There are no steps that request unrelated secrets, system configuration changes, or external data exfiltration. Note: producing file paths and line numbers is expected but will disclose any sensitive contents present in the repository.

Install Mechanism

No install spec and no code files — lowest-risk model. The skill is instruction-only and will not write files or install binaries itself.

Credentials

The skill declares no environment variables, no credentials, and no config paths. This is proportional to its purpose. The only potential runtime requirement is read access to the workspace and availability of git if the agent follows the '查看 git log' step.

Persistence & Privilege

always is false and the skill does not request elevated or persistent privileges. It does not modify other skills or system settings. Note that autonomous invocation is allowed by default (platform default) — combine with your usual governance if you restrict autonomous skills.

Assessment

This skill is an instruction-only read-only explorer for codebases and appears coherent and low-risk. Before installing, confirm you are comfortable with the agent having read access to the repository you analyze (it will inspect files, tests, and git history and may report file paths and line numbers). If you do not want git operations performed, ensure the agent environment lacks git or instruct the agent not to run git commands. No credentials are requested by the skill, but as with any codebase analysis, avoid running it on repositories containing secrets you don't want summarized or exposed.

探索模式 (Explore Mode)

进入只读深度探索模式，在回答前系统性地调研。

使用时机

✅ 使用此技能当：

• "帮我了解一下这个项目的架构"
• "这个功能是怎么实现的？"
• "调查一下为什么会出这个 bug"
• 需要全面理解一个不熟悉的代码库或系统

❌ 不要使用当：

• 已经很了解要查的内容
• 用户只是问一个简单的事实性问题

探索流程

1. 确定探索目标

• 要回答什么问题？
• 探索的范围是什么？（哪些目录/文件/系统）
• 需要达到什么深度？

2. 广度优先扫描

• 先看目录结构，建立全局认知
• 读 README、配置文件、入口文件
• 识别主要模块和它们的关系

3. 深度定向调研

• 针对目标问题，深入相关模块
• 跟踪关键调用链
• 查看测试用例理解预期行为
• 查看 git log 了解变更历史

4. 输出探索报告

结构化输出：

## 探索报告: [主题]

### 概述
一段话总结发现

### 架构/实现
- 关键文件和它们的职责
- 核心数据流/调用链
- 重要的设计决策

### 发现
- 发现1: ...
- 发现2: ...

### 回答
[针对原始问题的直接回答]

### 建议（可选）
如果发现了问题或改进空间

核心原则

• 只读不改 — 探索阶段不修改任何文件
• 系统性 — 不要只看一个文件就下结论，要交叉验证
• 记录路径 — 给出具体的文件路径和行号，方便用户定位
• 区分事实和推测 — 明确标注哪些是从代码看到的，哪些是推断的