design-increment ยท diff
git:20260918.4a145ae to git:20260918.c9794ad
4 added, 4 removed. Audit A to A.
---
name: design-increment
- description: Turn a project goal, options and research into one scoped implementable increment and an old-versus-new evaluation plan.
+ description: Turn a goal, options and research into one implementable increment and an old/new evaluation plan.
disable-model-invocation: true
---
- Use the [shared handoff contract](../../references/design-loop-contract.md). Inputs are request context, baseline, options and research evidence, plus any prior observations; equivalent caller-supplied material is sufficient.
+ Use the [handoff contract](../../references/design-loop-contract.md). Inputs: request context, baseline, options, research evidence and any prior observations; equivalent caller material suffices.
- For the named assignment `design-increment`, select one coherent increment, explain the decision and rejected alternatives, and specify changed behavior, boundaries, implementation outline, risks and required dependencies. The first cycle must target a minimum working PoC from the actual baseline.
+ For assignment `design-increment`, choose one coherent increment; explain the choice and rejected alternatives. Specify changed behavior, boundaries, implementation outline, risks and dependencies. The first cycle must target a minimum working PoC from the actual baseline.
- Return a design that an implementer can follow and an evaluation plan fixed before implementation: acceptance criteria, baseline behaviors to preserve, shared inputs/harness and conditions, expected old-state deficits, improvement measures and required versus optional checks. Mark any unresolved question that blocks implementation. If no change is justified, return that conclusion and its evidence for analysis rather than fabricate scope.
+ Return an implementable design and evaluation plan fixed before implementation: acceptance criteria, baseline behavior to preserve, shared inputs/harness and conditions, expected old-state deficits, improvement measures, and required/optional checks. Mark blocking questions. If no change is justified, return that conclusion and evidence for analysis.