How to use the Constraint Mapper agent skill (ljg-constraint)

What the skill makes the agent do

The skill asks one question: what few constraints really define this thing, and what do they make possible or impossible? The agent names those constraints and sorts each one into a level of hardness. A hard constraint is a world limit, where breaking it makes the system fail. A soft constraint is a rule with a cost for breaking it, such as law, licences, contracts or industry norms. A self-imposed constraint is a belief that people treat as fixed but could be redefined.

From that set of constraints, the agent works out the solution space they leave open, and uses it to explain the behaviour you observe. It is most useful when two sides of a debate share a goal but work from different default assumptions, and when you suspect that an old explanation is being treated as a fact.

The skill's own source describes the output as a Chinese prose essay, saved as a Markdown file. It does not run scripts or fetch anything from the internet. Its routing notes say that it is not meant for finding the generative forces behind a view (use ljg-rank for that), for drilling deep into one idea (ljg-think), for breaking down a book (ljg-book), or for dissecting a single concept (ljg-learn).

Install the skill from a ZIP file

Download the ZIP for skills/ljg-constraint from our page. Unzip it so that the folder containing SKILL.md sits inside the skills directory. For a skill available to all your projects, use ~/.claude/skills/. For a skill limited to one project, put it in that project's .claude/skills/ folder.

If you use the Claude apps instead of the file system, you can upload the skill under Settings, then Capabilities, then Skills. The source does not list any other installation method, and it lists no packages or extra dependencies to install.

• Personal use across projects: ~/.claude/skills/
• Single project: .claude/skills/ inside that project
• Claude apps: Settings, then Capabilities, then Skills, then upload

Trigger the skill with natural requests

The skill activates when your request matches its trigger words. You can use Chinese or English phrases. The source lists these triggers: 约束 (constraint), 找约束 (find the constraints), 这个领域的约束是什么 (what are the constraints of this field), 本质约束 (essential constraints), 什么框住了它 (what frames it), 硬约束还是自设约束 (hard or self-imposed constraint), 边界在哪 (where is the boundary), 这行为什么这么做 (why does this behaviour happen), constraint, and /ljg-constraint.

You can also just name a domain, profession, role, product or debate and ask for its defining constraints and solution space. The skill is written to respond to that kind of request as well.

• 约束 / 找约束: find the constraints
• 这个领域的约束是什么: what are the constraints of this field
• 这行为什么这么做: why does this behaviour happen
• 硬约束还是自设约束: is this a hard or a self-imposed constraint
• /ljg-constraint

Examples of requests and what the agent does

Ask about a role: "投资经理的约束是什么？" (What are the constraints of an investment manager?). The agent looks for constraints such as fund duration, trust from limited partners, information asymmetry and the split between upside and downside. It then shows how these push a manager toward chasing trends, waiting for consensus, or avoiding visible mistakes. The result is saved as a Markdown file.

Ask about a debate: "为什么产品和增长总吵架？" (Why do product and growth keep arguing?). The agent first writes out each side's default constraint, for example "do not disturb users" against "conversion must rise". It shows that both sides think they are arguing about a solution when they are working on different problems, and explains why each plan is rational inside its own constraints.

Test an assumed boundary: "这个行业必须重销售吗？" (Does this industry have to rely on heavy sales?). The agent asks whether this is a world constraint, a rule constraint or an industry explanation. It checks whether other eras, regions or players have got past it. If they have, the agent downgrades it to a soft or self-imposed constraint.

What the skill needs and how it stays safe

The source lists no requirements. It does not run scripts and does not use the network. The package has one file, SKILL.md, which holds the instructions. Because nothing runs and nothing is fetched, the skill's effect is limited to the analysis text the agent writes for you.

The skill is designed to describe before it advises. It keeps the diagnosis separate from advice and offers at most one line on which constraint might be loosened. Its gotchas also say not to label ordinary difficulties as constraints, and not to move habits or consensus into the hard-constraint category without testing them.

• Requires: nothing listed
• Runs scripts: no
• Uses network: no
• Files: SKILL.md

Tips for getting useful output

Give the goal together with the context in which it must be pursued, not just the goal. The skill asks you to complete the problem statement, and a clear problem produces a clearer set of constraints.

If your input is a dispute, name the sides and what each one takes for granted. The skill is built to show whether the sides are actually disagreeing about the same problem.

Expect the output to be a long reflective essay. If you want a quick fact, a step-by-step how-to, or a plan to beat the constraints, pick a different tool. The skill deliberately does not produce that.

• Include the goal and the conditions it must meet
• Name both sides of a debate and their default assumptions
• Use it for understanding, not for quick answers