One-page experiment contract

Copy this template after a research scoping call. Keep the first experiment small enough to complete, inspect, and stop without committing to a broader integration.

Working title

One sentence naming the capability or handoff question.

Participants and intended contributions

Participant	Intended contribution	Public credit if accepted
Name or stable alias	Protocol, artifact, model run, analysis, validation, task, documentation, or code	Repository profile, artifact, or contribution record

This table records intent, not contributor or paper-authorship status. Update it to reflect work actually accepted.

Research-to-production decision

What decision can the current functional evaluation not support?

Falsifiable question

State one question whose answer can be yes, no, or unknown under the declared evidence. Avoid a general claim about model or production quality.

Smallest public representation

• Public, redistributable, or synthetic input:
• License or permission basis:
• SV-Gap track: generation, diagnosis, repair, benchmark adapter, taskpack, or checker comparison:
• Why the artifact represents the handoff question:

Frozen protocol

• Task or taskpack identifier and digest:
• Model/configuration label or stable alias, if applicable:
• Generation settings and repeat count:
• Functional evidence source:
• Declared clock/reset intent:
• Structural backend and version:
• Prespecified pass, fail, unknown, and tool-error handling:

Evidence and claim boundary

Production question	Evidence that can answer it	Evidence available now	Missing evidence

The result will establish:

The result will not establish:

Deliverable and stopping rule

• Smallest publishable evidence unit:
• Reproduction material:
• Stop after the first run when:
• Expand or replicate only when:

Owners and next decision

Action	Owner	Target date
Draft/freeze protocol
Provide or approve public artifact
Run and validate
Review interpretation

Next decision: go, revise, out of scope, or complete.