Problem
After option 3 ("Implement directly") is chosen from a next-steps menu and implementation completes, there is no agent-driven prompt about whether to run a follow-up review pass. The user is expected to remember to invoke review-change (or orchestrate-review) themselves, and to make the review-vs-no-review judgment unaided.
Whether a review pass is worth running depends on what was actually built — the agent has visibility into the diff, scope, and signals (e.g., complexity, plan source) that the menu moment lacks. A post-implementation prompt is the natural place for that recommendation.
Context
Follow-up to #544, which corrected the framing of option 3 in next-steps-after-plan.md but stopped short of adding the post-implementation prompt. The threshold table in _data/complexity-classification.md now anticipates two outcomes ("Implement directly" / "Implement directly with follow-up review") that no agent currently surfaces.
Acceptance
- After implementation, the user receives a recommendation about whether to run a review pass (and which variant if so).
Problem
After option 3 ("Implement directly") is chosen from a next-steps menu and implementation completes, there is no agent-driven prompt about whether to run a follow-up review pass. The user is expected to remember to invoke
review-change(ororchestrate-review) themselves, and to make the review-vs-no-review judgment unaided.Whether a review pass is worth running depends on what was actually built — the agent has visibility into the diff, scope, and signals (e.g., complexity, plan source) that the menu moment lacks. A post-implementation prompt is the natural place for that recommendation.
Context
Follow-up to #544, which corrected the framing of option 3 in
next-steps-after-plan.mdbut stopped short of adding the post-implementation prompt. The threshold table in_data/complexity-classification.mdnow anticipates two outcomes ("Implement directly" / "Implement directly with follow-up review") that no agent currently surfaces.Acceptance