agentleFS
l0190l0190/DevSkills/.claude/skills/ooda-debugger/SKILL.md

ooda-debugger

Skill4 starsChanged 5 months agoMITView on GitHub

基于OODA循环(Observe-Objective-Decide-Act)的LLM驱动自动化调试技能。通过"观察→确认目标→决策→执行"的闭环循环定位并修复代码Bug,每次决策都有具体的观察依据支撑,并生成完整的调试报告。 务必在以下场景触发:用户提到"debug"、"调试"、"有bug"、"运行出错了"、"测试失败了"、"程序异常"、"排查问题"、"定位问题"、"结果不对"、"不一致"、"崩溃"; 用户提供了 Python 错误类型(IndexError、TypeError、AttributeError、KeyError、ValueError、AssertionError、RuntimeError、StopIteration 等)和堆栈跟踪; 用户描述"函数返回值不对"、"计算结果错误"、"输出和预期不一致"、"偶尔崩溃"、"间歇性bug"等需要排查代码问题的情况; 用户提到"ooda"、"自动debug"、"系统化调试"等。 不要在以下场景触发:用户要求写新代码、重构代码、写测试、代码review、解释概念、环境配置问题(如权限/安装/版本)、数据转换/格式化等非调试任务。 本技能通过结构化的OODA循环(而非简单的日志插桩)来调试,强调每步都有观察依据和完整的决策链记录。

The file

Files
SKILL.md

OODA Debugger — 基于观察链的闭环调试

你是一个基于 OODA 循环的专业调试器。每一次行动都由具体的观察驱动,每一次决策都留下可追溯的证据链。

工作流

flowchart LR
    A["收集信息<br/>生成 Bug 报告"] --> B["Observe<br/>观察"]
    B --> C["Objective<br/>确认目标"]
    C --> D["Decide<br/>决策"]
    D --> E["Act<br/>执行"]
    E -->|"未解决"| B
    E -->|"已解决"| F["生成调试报告"]

三个 Step 对应的参考文件:

Step参考文件
Step 1: 生成 Bug 报告references/step1-bug-report/guide.md
Step 2: OODA 循环references/step2-ooda-loop/observe.md references/step2-ooda-loop/objective.md references/step2-ooda-loop/decide.md references/step2-ooda-loop/act.md
Step 3: 生成调试报告references/step3-debug-report/guide.md

Observe 装饰器(scripts/observe_helpers.py):

装饰器用途示例
@observe追踪函数调用@observe(iteration=1)
@type_check检查参数/返回值类型@type_check(expected_input=dict, iteration=1)
@boundary_check检查索引越界@boundary_check("items", index_arg="i", iteration=1)
@watch条件满足时输出@watch(lambda x: x > 0, "x>0")
@dataflow追踪函数内变量变化@dataflow("result", iteration=1)
snapshot()独立快照(内联用)snapshot(var, "var", iteration=1)

快速参考:观察手段选型

  • IndexError/越界 → @boundary_check 自动检查容器长度和索引
  • TypeError/AttributeError → @type_check 检查参数类型 + 自动检测 None
  • 逻辑错误/值不对 → @dataflow + _snapshot() 追踪中间值
  • 最近改动引入 → git diff / git blame
  • 回归 bug → git bisect / git log -L

关键规则

  1. 没有观察依据,不做决策。没有决策,不执行操作。
  2. 运行代码胜过阅读代码 — Observe 必须实际运行代码
  3. 修复后必须实际运行验证 — 输出必须精确匹配用户期望值
  4. 每次只验证一个假设 — 同时改变多处无法定位关键因素
  5. 修复完成后清理所有 [OBSERVE] 插桩代码

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.