mr-fang-vlsi/eda-hypothesis-experiment-designer

EDA Hypothesis Experiment Designer

When to use

Use this skill after idea selection and before method implementation or expensive experiment submission.

Knowledge And Tool Interaction

1. If the task needs shared KB or tool lookup before formalizing hypotheses, delegate that step to eda-context-accessor.
2. Use idea memos, KB context, and prior evidence artifacts as the basis for falsifiable hypothesis design instead of inventing hypotheses from scratch.
3. When related local experiments already exist, use eda-experiment-phenomenology-analyst artifacts so hypothesis design reflects repeated practical evidence rather than only theoretical intuition.
4. Write experiment-design outputs so downstream implementation and validation skills can consume them directly.

Outputs

1. hypothesis_experiment_matrix.tsv
2. experiment_design_note.md
3. promotion_gate.md

Hard rules

1. Each hypothesis must have explicit disproof condition.
2. Metrics must match decision question (not just easy-to-collect metrics).
3. Include at least one negative-control or ablation check when feasible.

Operational References

1. Load references/assumption-to-hypothesis.md when extracting assumptions from an idea memo and turning them into falsifiable statements.
2. Load references/matrix-rules.md when filling the experiment matrix columns, metrics, confounders, and pass/fail rules.
3. Load references/promotion-gate-design.md when defining the evidence needed to move from hypothesis to implementation.