Three parallel reads on N focused questions, with disagreement preserved.
Steps
Step 1 — Dispatch
- Reviewers: R0 (you, concurrent) + R1, R2 (two
general-purposesubagents in parallel). - Prompt (identical for all three): artifact path(s), the questions, a confidence (0.00–1.00) and impact score per recommendation.
- Preflight: if any question is vague, sharpen via
AskUserQuestionfirst. - Output schema per question: recommendation, confidence (0.00–1.00), impact — render
Label (value): Minimal (0.25) · Low (0.5) · Medium (1) · High (2) · Massive (3) — 1-2 sentence reason. - Unprompted observations (any reviewer): 1-sentence claim, raising reviewer(s), confidence 0.00–1.00, impact; merge duplicates.
Step 2 — Synthesize and confirm
- Sweep: for questions with max < 0.80 AND any reviewer < 0.70, run
second-opinionon the highest-confidence reviewer's pick and raise Max to the synthesized score if higher. - Order: sort by max confidence descending (post-sweep).
- Dependency override: if a decision can invalidate later questions, walk it first.
- Table:
| # | Question | R0 | R1 | R2 | Max | Impact |
- Low-consensus filter: drop questions where post-sweep max < 0.60 AND no reviewer ≥ 0.75; list in Step 5.
- Checkpoint: use
AskUserQuestionto confirm the walk order before walking.
Step 3 — Walk questions one at a time
Queue rule: drop later questions invalidated by an approved decision, with a one-line reason.
For each question:
- Present: restate the question, quote the artifact span if location-specific, and show R0/R1/R2 side-by-side (plus second-opinion's synthesized recommendation, if Sweep ran).
- Decide:
AskUserQuestionwith options: apply / alternative / defer. Apply if actionable; otherwise record.- On pushback: run
second-opinionon the challenged pick.
- On pushback: run
Step 4 — Walk unprompted observations
List all observations. Walk those with confidence ≥ 0.70 using Step 3's sub-procedure (skip R0/R1/R2 split).
Step 5 — Summary
One line each:
- Decisions applied — question # + pick.
- Decisions deferred — reason.
- Questions skipped (upstream invalidation) — reason.
- Questions dropped (low panel consensus) — question + max score.
- Unprompted observations — applied / deferred / dropped.