hughyau avatar

Investigation First

This stops you from making decisions before you understand what's actually happening. It kicks in when you're about to give advice but realize you don't have enough context, evidence, or firsthand information. You'll output a clear investigation checklist before doing anything else: what files to read, what state to check, what constraints to verify. The structure is borrowed from Mao's "No investigation, no right to speak," which is more practical than it sounds. It forces a separation between gathering facts and drawing conclusions, and won't let you work backwards from a hunch. Best for unfamiliar codebases or when you catch yourself pattern-matching instead of looking at what's really there.

skills.sh installs
305
GitHub stars
3.9K
last updated
Aug 21, 2026

Install

$ npx -y skills add hughyau/qiushi-skill --skill investigation-first --agent claude-code

Installs into .claude/skills of the current project.

Files

调查研究

"没有调查,没有发言权。" —— 《反对本本主义》

调查研究是落实「实事求是」的入口:不是为既定结论找材料,而是让事实规定判断。

用 / 不用

用:

  • 要做决策,但对实际情况了解不够
  • 面对陌生的代码库、系统或领域
  • 别人的方案听起来合理,但你不清楚其适用条件
  • 发现自己在用"应该是""一般来说"作答

不用:

  • 已经调查过,现在是执行阶段
  • 用户已提供完整上下文和决策依据
  • 答案可以直接跑一下验证,跑比查快

操作规程

  1. 一句话写出调查目的:「我要弄清楚的是:……」。没有目的的调查是漫游。
  2. 列调查提纲(checkbox),覆盖四个面:
    • 现状:现在是什么样(读源码、配置、数据,不只看文档)
    • 历史:为什么变成这样(git log、blame、变更记录)
    • 关联:牵涉哪些约束(接口、依赖、调用方、部署环境)
    • 多方:有哪些不同说法(文档 vs 代码 vs 运行结果 vs 用户描述)
  3. 亲自接触第一手材料:直接读、直接跑、直接看日志。二手总结只能作线索,不能作证据。
  4. 边查边记,分三栏:每条发现标注来源;把"看到的"和"推出来的"分开。
  5. 输出调查结论(模板见下)。结论必须先于任何行动方案;不得边查边给方案。
  6. 事实与预期不符时,改判断,不改事实。查不清的,诚实写"未知"。

输出模板

调查目的:……
调查范围:读了 …… / 跑了 …… / 问了 ……

已核实的事实(附来源):
- ……(来源:文件:行 / 命令输出 / 日志)
合理推断(说明推理链):
- ……(因为 …… 所以 ……)
未知 / 存疑:
- ……(需要 …… 才能确认)

调查结论:
- 现状是:……
- 关键约束是:……
- 我原来不知道、现在知道的是:……
- 基于以上,我的判断是:……(置信度:高 / 中 / 低)

纪律

  • 禁止在"未知"栏非空且关系到结论时下定论:要么补查,要么把不确定性写进结论。
  • 禁止走马观花:只看目录结构和 README 就总结的,不算调查。
  • 调查有预算:先写提纲,按提纲查完即止;发现新的大问题时另开一轮,而不是无限扩张。

交接

  • 调查揭示多头矛盾、抓不住重点 → contradiction-analysis
  • 结论已形成、需要验证 → practice-cognition
  • 范围大、可分头查 → 派遣 investigator 子 agent 并行调查,主线只汇总三栏

原著依据:original-texts.md

Sponsors

Details

Repository
github.com/hughyau/qiushi-skill
First seen
Jun 3, 2026

More from hughyau/qiushi-skill

All 11

Recommended

1juliusbrussee avatarcavemanjuliusbrussee/cavemanUltra-compressed communication mode. Cuts output tokens 65% (measured) by speaking like caveman while keeping full technical accuracy. Supports intensity levels618.3Kskills.sh installs
2mattpocock avatargrill-memattpocock/skillsA relentless interview to sharpen a plan or design.1.5Mskills.sh installs
3shadcn avatarimproveshadcn/improveSurvey any codebase as a senior advisor and produce prioritized, self-contained implementation plans for OTHER models/agents to execute. Strictly read-only on s40Kskills.sh installs
4obra avatarsystematic-debuggingobra/superpowersUse when encountering any bug, test failure, or unexpected behavior, before proposing fixes326.7Kskills.sh installs
5forrestchang avatarkarpathy-guidelinesforrestchang/andrej-karpathy-skillsBehavioral guidelines to reduce common LLM coding mistakes through explicit assumptions, simplicity, and verifiable success criteria.17.9Kskills.sh installs
6vercel-labs avatarfind-skillsvercel-labs/skillsHelps users discover and install agent skills when they ask questions like "how do I do X", "find a skill for X", "is there a skill that can...", or express int4.3Mskills.sh installs