Guide · Agent skill
How to use the ljg-structure agent skill: find hidden structure
Updated · written with AI from the project's own documentation and checked against our catalogue data
ljg-structure is an agent skill, a folder of instructions that a Claude agent loads when your request matches its trigger phrases. It is built for problems that keep repeating, such as a team that always gets derailed the same way, and it looks for the cause-and-effect pattern underneath. After this guide you will know how to install it, how to phrase a request so it triggers, and how to read the result.
Structure Wind-Tunnel Analysislijigang/ljg-skills · MIT Download the skill Quick start
- 1Download the ZIP. Download the skill ZIP from this page. It contains SKILL.md and Workflows/FindStructure.md.
- 2Unzip into a skills folder. Unzip it into ~/.claude/skills/ for personal use, or into the .claude/skills/ folder of a project.
- 3Or upload in the Claude apps. In the Claude apps, open Settings, then Capabilities, then Skills, and upload the skill.
- 4Describe a recurring problem. Write a concrete situation that keeps repeating, using a trigger phrase such as 找结构 or 结构风洞.
- 5Review the motif and structure. Check the motif, the example it is anchored to, and the relationship diagram for whether they match what you saw.
- 6Read the conclusion. Look at the changed condition, the transferable rule, and the note on where the model might be wrong.
- 7Keep or discard the saved file. If it saved a file, find it under ~/Context and keep it if useful, since the source describes it as a Markdown note.
What the skill makes the agent do
Given a piece of information, the skill looks past surface details for a recurring type of dilemma, called a motif in the source. It then draws the specific relationships that make the pattern work, and keeps a concrete example that the rest of the analysis hangs on.
It tests the idea by changing one key condition inside its analysis to see where the prediction changes and where the model stops working. The final answer is a short conclusion with a transferable rule and a way to recognise the pattern in the real situation. The skill is designed to keep the output light: one clear motif, one relationship diagram, and one conclusion.
- Separates four ideas: topic (what it is about), motif (the recurring dilemma), analogy (a lead, not proof) and structure (the cause-and-effect skeleton).
- Keeps one example as an anchor instead of listing many cross-domain cases.
- Does not show the wind-tunnel steps as a separate section; they stay internal.
Install the skill from the ZIP
Download the ZIP for this skill from this page and unzip it. The skill folder contains SKILL.md and a Workflows/FindStructure.md file. Put the folder where your agent can read it: either your personal skills folder or the .claude/skills/ folder inside a specific project.
If you use the Claude apps instead of a local setup, you can upload the skill under Settings, then Capabilities, then Skills.
- Personal install: unzip into ~/.claude/skills/
- Project install: unzip into the .claude/skills/ folder of your project
- Claude apps: Settings → Capabilities → Skills, then upload the skill
Trigger the skill with natural requests
The skill loads when your request matches its trigger phrases. The source lists phrases such as 找结构, 母题是什么, 结构风洞, 背后的结构 and 不要只做AB测试. You can also ask for the generative mechanism beneath a problem in your own words.
Name the messy situation and say what you want. A clear request gives the skill something to analyse; a bare domain name without a described tension does not.
- 找结构：为什么我们的项目评审总在同一个地方卡住？
- 母题是什么：每次上线前的沟通总是临时补救，背后是什么结构？
- 不要只做AB测试：我们的新功能数据波动，帮我看背后的机制。
- Use 结构风洞 to ask for the analysis with the wind-tunnel test built in.
Read and save the result
The default output is in Chinese and roughly 800 to 1500 characters. It uses four top-level headings: 输入, 母题, 结构, and 结论. The 结论 section first describes the key difference when one condition changes, then states a transferable rule, then points back to the original situation.
By default the skill saves the result as a Markdown file under ~/Context, named with a timestamp and the topic. If you write 只分析, 不落盘, or read-only, it gives the same analysis in the chat without creating a file.
- File pattern: ~/Context/{timestamp}--结构-{topic}__structure.md
- The file uses Markdown with a short front matter block, not Org format.
- After saving, the skill reads the file back and reports its path.
Safety notes: scripts and network
The source states that this skill does not run scripts and does not use the network. Its declared requirements list nothing. Its only side effect by default is writing one Markdown file into ~/Context, unless you ask for read-only analysis.
Because it writes files, check the saved path if you care where your notes go. Because it is a reasoning method, treat its output as a hypothesis to test against your own evidence.
- Runs scripts: no
- Uses network: no
- Requires: nothing listed
- Default file write: ~/Context/ Markdown file
Tips for better results
Describe the concrete situation: who did what, what changed, and when the decision went wrong. The skill needs at least one observable detail to work from. If you only give an abstract claim, it will add an illustrative example and say so.
Ask for one question at a time. The skill is built to find one dominant motif, so a focused problem works better than a long list of issues.
- Include one observable moment, such as a meeting where the first speaker set the frame.
- Use 只分析 if you do not want a file written.
- Do not use it for simple fact lookups or one-step tasks with a clear answer.
What people use it for
- Finding out why a meeting or team keeps returning to the same problem.
- Building a conceptual model to test, rather than only summarising a document.
- Explaining a messy situation with more structure than a standard A/B test explanation.
- Writing a short, saved Markdown note of a structural analysis for later reference.
Frequently asked questions
How do I install the ljg-structure skill?
Download the ZIP from this page and unzip it into ~/.claude/skills/ for personal use, or into a project's .claude/skills/ folder. In the Claude apps, you can instead upload it under Settings, then Capabilities, then Skills.
Does ljg-structure require anything else to run?
The source lists no requirements. It does not run scripts and does not use the network.
Is ljg-structure free to use?
It is released under the MIT licence, as stated on this page. The source does not mention pricing.
What is the difference between this skill and a normal summary?
A summary restates what a text says. This skill looks for the recurring dilemma underneath, the cause-and-effect relationships that produce it, and where the explanation breaks. The source says it is not meant for plain summaries.
Can ljg-structure be used for medical or legal decisions?
The source says analogies from this skill can generate hypotheses in high-stakes medical, legal or safety areas, but they cannot replace domain evidence.
Why did the skill not trigger on my request?
Trigger phrases such as 找结构, 母题是什么, 结构风洞 and 背后的结构 are listed in the skill. Requests that are simple fact lookups or single-step tasks with one clear answer are explicitly outside its scope.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
Use ljg-structure if you keep meeting the same problem and want a testable model of why it recurs, or if you want more than a standard A/B explanation of a messy situation. Pick something else for simple fact lookups, summaries, or tasks with one clear answer, and never rely on it alone for high-stakes medical, legal or safety decisions.
Download the skill