four-forms
Skill1 starsChanged 6 days agoApache-2.0View on GitHub
Think a subject through with the Four Forms of Thinking (by GoodRelax, compiled from classic thinking moves). The four forms, in order, are Divide, Compare, Infer, and Transform; each has two core moves and one optional move. Reach for it when a person is stuck on a subject, a plan or a business idea, and wants to think it through. Not for a bug or a coding task. It works in the chat, and on the local whiteboard of this plugin (server board-control) when the person asks for it or agrees to it. Use when the user types /gr-whiteboard:four-forms, asks to think a subject through with the Four Forms of Thinking, or when consult-on-board or issue-driven picks it.
In this repository5 other files it gives its agents.
- consult-on-boardSkill
- debug-board-serverSkill
- five-whysSkill
- issue-drivenSkill
- talk-on-boardSkill
The file
We split this skill into the rules it gives an agent
3 rules. About 0 are common, 3 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈3
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 25
Start with Lv1 Divide: define the words.
- Distinctive: line 36
Check the facts of that field.
- Distinctive: line 39
Ask the person for the case most likely to break the match.
≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
four-forms
Use the Four Forms of Thinking and the board to rebuild how the person looks at a subject, from a higher vantage point, one form at a time. Use them when the person has run out of ideas, is stuck, wants a fresh idea, or wants to compare one thing with another.
Talk in the chat first, one form at a time in short bullets. Open the board only when the person asks for it, or when a picture would make it clearer and you ask "Shall we sort this out on the whiteboard?" and they agree; then run it as the skill talk-on-board says. After open_board, start the watch Monitor of talk-on-board (Open step 2, client.js watch) at once, or you miss every Send. Draw the person's words, not your own ideas, and validate them, as consult-on-board says.
The forms
| Form | Core moves | Optional move | It holds when |
| Lv1 Divide | Divide (whole into parts), Name (a sign for each) | Classify (parts into kinds) | The parts do not overlap and make the whole. One rule sorts the parts into kinds, each part in one kind. One sign for each object. Name a part by what it does or asks of the person, not by what it is made of: a name of that kind carries over to another field |
| Lv2 Compare | Analogy (alike), Contrast (opposite) | Orthogonal (unrelated) | Analogy points the same way, Contrast the opposite way, Orthogonal shows no relation. Analogy keeps the relations between the parts, and it may cross into a far field |
| Lv3 Infer | Induction (find the rule), Deduction (find the result) | Abduction (find the cause) | Induction: a law from two or more cases (conjecture). Deduction: a result from a cause and a law (certain). Abduction: the cause that best explains the result (selection) |
| Lv4 Transform | Concrete (add attributes), Abstract (remove attributes) | Synthesis (recombine abstractions) | Concrete adds attributes and Abstract removes them. Synthesis recombines several abstractions with weights into one. Abstract keeps the relations that matter and stops while they still say something: a shape that fits anything is too high |
Run
- Show the whole flow first:
draw_mermaidwithfileset toSKILL_DIR/data/forms-ja.mmdfor a Japanese speaker, orSKILL_DIR/data/forms-en.mmd(absolute paths; SKILL_DIR is the folder of this file). Above it, write the title withdraw_skeletonsandat: the framework name and the subject (e.g. "Four Forms of Thinking Sample Inc. new product"). Below it, write "Gray = optional" in gray. - After the drawing, make the optional moves gray (Classify, Orthogonal, Abduction, Synthesis: the box and its text,
edit_itemscolor gray), and make the frames of the three link boxes between the forms transparent (strokeColor"transparent"). Find the ids withread_board(typetext gives each text with its box incontainerId). - Start with Lv1 Divide: define the words. If the person starts at Lv2 or later, skip Lv1, and come back to it when a term needs a definition.
- Go one form at a time. Use the optional moves only when they help. Write the person's words in a column to the right of the flow, one heading per form, and pass
attowrite_answerso that your answers stay beside that column. - You may go back to an earlier form when a later one shows a gap.
Far field
When it would help, or the person asks for it: at Lv2 Analogy, after the person has compared the parts of their own subject, say the shape in one line, in relations only, with the words of the person's field taken out. If the line fits anything, it is too abstract: put one relation back.
Then offer once, as an invitation with examples, in words like: "The same shape may show up far from your field: for example A, B, or C. Which of these might work? Or does another one come to mind?" Give two or three fields that have nothing to do with the person's field and little to do with each other, each with a few words on why it might fit. They are seeds, not answers: the person may pick one, change one, bring their own, or pass. If the person picks one, or asks for it, do this. Here you may bring your own ideas: the candidate fields and the facts about them are yours, and what they mean at home is the person's.
- Think the chosen field through with the person: what happens there, and who watches what.
- Check the facts of that field; say so when you are unsure.
- Lay out the pairs: here on one side, there on the other, each relation beside its partner. On the board, use
draw_tablewith two columns (this is not a frame) and put it next to the Lv2 notes. In the chat, use a two-column list. - Say which relations do not carry over. Every far field has some; name at least one, and name the part of the person's problem it says nothing about.
- Ask the person for the case most likely to break the match. When it breaks, go back to the far field and ask how that field handles the same case. A match improves in both directions.
- Hand it back at Lv4 Concrete: the person says what the far-field practice becomes at home, and how they will test it there.
Stay out of a far field that only repeats the person's own field in other words.
A far field gives ideas, not proof. Test them in the person's own field.
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
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.

