innovation-breakthrough-design

Files (3)

• SKILL.md
• breakthrough-history.md · references/
• innovation-card.md · templates/

SKILL.md (first lines)

# Innovation and Breakthrough Design

Use this skill when the user wants genuinely new directions, not merely a summary or incremental variant. Treat creativity as a search process constrained by mechanisms, contradictions, and experiments.

## Breakthrough pipeline

1. **Define the opportunity.** State the user, unmet need, target outcome, constraints, and why current approaches fail.
2. **Decompose the frontier.** Map the system into functions, bottlenecks, invariants, interfaces, resources, and failure modes. Search adjacent domains for mechanisms that solve an analogous bottleneck.
3. **Extract the pattern.** Compare historical breakthroughs as sequences of anomaly, reframing, mechanism, enabling tool, prototype, and validation. Read [breakthrough-history.md](references/breakthrough-history.md) for prompts, not as a guarantee of novelty.
4. **Generate structured candidates.** Create multiple candidates using analogy transfer, inversion, composition, constraint relaxation, scale change, and representation change. For equations, specify objects, dimensions, symmetries, boundary conditions, and limiting cases.
5. **Run a candidate tournament.** Compare candidates against the same baseline and score mechanism clarity, preserved constraints, resource/energy/material accounting, evidence, reversibility, learning value, and falsifiability. Do not rank a candidate before its missing inputs are visible.
6. **Attack the candidates.** Test prior-art boundary, internal consistency, edge cases, safety, adoption friction, rival explanations, and cost of error. Kill weak ideas early.
7. **Prototype the smallest proof of value.** Select a bounded experiment, simulation, derivation, mockup, or user test with a clear success threshold, negative control, and decision-reversal condition.
8. **Iterate and communicate.** Convert the best candidate into an innovation card with mechanism, assumptions, baseline, test, risks, proof gaps, and next experiment.

## Candidate quality gate

A breakthrough candidate must identify a mechanism, not only a desirable outcome. It must make a non-obvious connection while preserving relevant constraints, state what could falsify it, and expose its nearest prior art or analogy. Never claim legal novelty, scientific truth, or commercial viability without a dedicated search and validation.

## Default output

Use `templates/innovation-card.md`. Provide at least three candidates before selecting one, unless the user explicitly asks for a single constrained design.

From minagayid/innovator, MIT.

More from minagayid/innovator