Thank you for your interest in improving The Cube. This is an open-source thinking tool, and contributions from diverse perspectives make it better.
The most valuable contribution is real-world feedback. After running a Cube analysis, use /cube-feedback to capture structured feedback. This helps us understand:
- Which positions produce the most valuable insights
- Which positions feel redundant
- What angles are missing
- Whether the analysis length is right
To share your feedback with the project:
- Open a GitHub issue with the
feedbacklabel - Include your feedback file (from
.cube/feedback/) - Redact any sensitive problem details -- we care about which positions were valuable, not the specifics of your problem
If The Cube produces output that is malformed, inconsistent, or fails to follow the protocol:
- Open a GitHub issue with the
buglabel - Describe which skill you used (
/cube,/cube-quick, etc.) - Include the relevant portion of the output
- Note your Claude Code version if possible
If you have ideas for improving how a specific framework is applied:
- Open a GitHub issue with the
frameworklabel - Specify which face/framework you are addressing
- Describe the current behavior and the improvement
- If possible, include a before/after example showing how the improved framework produces better insight on a real problem
If you believe a specific combination of frameworks is missing or underserved:
- Open a GitHub issue with the
rotationlabel - Describe the combination (which frameworks, and why this combination produces something the existing rotation protocol misses)
- Include an example of the insight this combination would produce
Documentation improvements are always welcome:
- Fix typos, clarify explanations, improve examples
- Add translations
- Create tutorials or guides
- Submit as a standard pull request
If you want to create a new domain-specific cube (e.g., Legal, Design, Leadership, Finance):
- Open a GitHub issue with the
stack-proposallabel - See "Stack Proposals" section below for full requirements
If you have an idea for a new /cube-* slash command:
- Open a GitHub issue with the
skilllabel - Describe the use case: when would someone use this command?
- Describe the expected input and output
- Explain how it relates to the existing skill suite
Skills are Markdown files in .claude/skills/. Each skill requires:
---
name: [skill-name]
description: "[one-line description visible to users in skill listings]"
---
<skill-name>
[skill content -- instructions for Claude to execute]
</skill-name>- Be directive. The skill file is an instruction set, not documentation. Tell Claude what to do, step by step.
- Include self-prompts. For each analysis point, provide the specific question Claude should ask itself. "What are the second-order effects?" is better than "think about consequences."
- Specify output format. Include a template showing exactly how the output should be structured.
- Include execution notes. Common failure modes and how to avoid them.
- Test with real problems. Before submitting, run the skill on 2-3 real problems and verify the output is specific and valuable.
- Specificity: Every instruction must push toward problem-specific output, not generic advice
- Geometry consistency: All references to the cube (faces, edges, axes, corners) must use the correct geometric relationships
- Self-contained: Each skill file must work independently -- no external dependencies
- Backward compatible: Changes should not break existing workflows
Follow this format:
[component] short description
Longer description if needed.
Examples:
[cube] add cross-reference protocol to edge rotations
[cube-quick] fix position numbering in axis tensions
[docs] clarify installation instructions for global install
[cube-face] add 4-phase protocol for boundary conditions
Proposing a new stack (a new domain-specific cube) is a high-impact contribution.
-
Domain specificity: The stack must serve a domain where the Core cube's general-purpose frameworks miss domain-specific angles that practitioners would recognize as essential.
-
Six orthogonal frameworks: Each framework must see something the other five cannot. No two should be reducible to each other.
-
Three productive opposite-pair tensions: Each pair of opposite faces must create genuine dialectical tension, not artificial grouping. Describe each tension.
-
Named rotations: All 12 edges, 3 axes, and 8 corners must have meaningful names that capture the character of each combination.
-
Universal applicability within the domain: Every framework must work across the full range of problems in that domain. A sales framework that only works for SaaS does not belong.
-
Demonstration: Provide a worked example showing the stack applied to a real problem, highlighting at least 3 rotations that produce insights the Core cube would miss.
- Open a GitHub issue with the
stack-proposallabel - Address all six requirements above
- Community discussion (minimum 2 weeks)
- If consensus supports it, the proposer creates the full skill file
- Review and merge
| Stack | Command | Domain |
|---|---|---|
| Core | /cube |
General problem solving |
| Sales | /cube-sales |
Sales methodology |
| Marketing | /cube-marketing |
Marketing strategy |
| Coding | /cube-coding |
Engineering problem solving |
Proposing a new framework (a new "face" of an existing cube) is the highest-impact contribution and requires the highest bar.
-
Orthogonality: The framework must see something genuinely different from all existing frameworks. If it can be reduced to a combination of existing frameworks, it belongs as an edge or corner rotation, not a new face.
-
Universal applicability: The framework must work on technical, business, personal, creative, and strategic problems. Domain-specific frameworks do not belong on the cube.
-
Adversarial to framing: The framework must force the user out of the problem's native frame. If it reinforces the existing framing, it does not meet The Cube's core design principle.
-
Productive tension pairing: The framework must create a meaningful dialectical tension when paired with at least one existing framework as an opposite face.
-
Demonstration: Include a worked example showing the framework applied to a real problem, with clear evidence that it produces insight the existing six frameworks would miss.
- Open a GitHub issue with the
framework-proposallabel - Address all five requirements above
- Community discussion (minimum 2 weeks)
- If consensus supports it, a prototype skill is created and tested
- If testing validates it, it is merged into the protocol
Note: Adding a seventh framework would require restructuring the cube geometry. This is possible (e.g., moving to a different polyhedron or adding a "floating" framework), but it is a significant protocol change. Propose only if the value clearly justifies the complexity.
- Be specific. Vague feedback is hard to act on. "I didn't like the edge rotations" is less useful than "The Autopsy (Position 7) repeated what Inversion and Pre-Mortem already said instead of finding the intersection."
- Be constructive. Criticism of the protocol is welcome. Criticism of people is not.
- Show your work. Claims about framework improvements should include examples. "This would be better" needs to show how it would be better on a real problem.
- Respect the geometry. The cube structure is not arbitrary. Before proposing changes to the rotation protocol, understand why the current structure exists (see ETHOS.md).
By contributing to The Cube, you agree that your contributions will be licensed under the MIT License.