Define a bounded-brief ablation protocol

[nothingnesses]

This issue records a proposal. Implementation requires a selected work brief and separate execution approval.

Background and problem

D-068 retains the unresolved question of which prompt content improves correct next actions per token. The old ledger-output experiment does not fit the retained bounded-brief direction.

Current next output is a state projection. The proposed per-role brief command is not shipped.

Proposal

Prepare a fixed protocol for a later comparison of bounded brief variants.

Define expected actions and constraints independently of the candidate text. Compare:

• Correctness.
• Important constraint violations.
• Token cost.

Scope and exclusions

Exclude:

• Implementation of brief.
• Selection of its schema.
• Restoration of the ledger experiment.
  The command-contract proposal covers that product direction.

Do not claim that shorter output improves performance without task evidence.

Acceptance criteria

• Candidate variants and comparison rules are explicit before execution.
• Expected outcomes do not derive from the candidate answer.
• The protocol identifies prompt versions and token measurements.
• Incorrect actions and omitted constraints remain visible.
• The proposed conclusion states sample and task limits.
• Budget and execution scope require separate approval.

Alternatives and open questions

A proposed brief can support a protocol proof, but cannot establish shipped-command effectiveness. Prefer a stable candidate supplied by the architecture work before execution.

References

• Decision report, D-068 and the status/next/brief direction.
• current workflow documentation.
• D-068.