| name | hypothesis |
| description | Guides hypothesis-driven investigation for Parallax development. Invoke before starting any investigation, feature, or experiment. |
/hypothesis -- Hypothesis Workflow
Guides hypothesis-driven investigation for Parallax development.
When to Use
Invoke /hypothesis before starting any new investigation, feature, bug fix, or experiment on the Parallax codebase.
Protocol
- State your hypothesis. What do you expect, and why? Be specific and falsifiable.
- AI generates alternatives. Propose 2-3 alternative hypotheses for consideration.
- Human selects. Pick the hypothesis to test (may be the original or an alternative).
- Define success/failure criteria. What measurable outcomes support or refute the hypothesis?
- Design and run the test. Implement the experiment, collect results.
- Record the conclusion. Supported, refuted, or inconclusive -- with evidence. Document regardless of outcome.
Output Format
## Hypothesis: [descriptive name]
**Statement:** [specific, falsifiable claim]
**Rationale:** [why you expect this]
**Success criteria:** [measurable outcome if supported]
**Failure criteria:** [measurable outcome if refuted]
**Status:** proposed | testing | supported | refuted | inconclusive
**Evidence:** [results, data, references]
Rules
- Never skip the hypothesis step. No implementation without a stated expectation.
- Negative results are valuable. Document them with equal rigor.
- If inconclusive, state what additional evidence would resolve it.
- Reference @CONSTITUTION.md principles 1, 3, 6 when documenting.
- Record hypotheses in session summaries (docs/sessions/) for continuity.
