review-feedback-orchestrator

# Review Feedback Orchestrator ## Inputs - Review feedback source (one of): - comments earlier in the conversation, - pasted review text, - markdown file path. - Optional scope hints (branch, folders, files). ## Workflow 1. Collect and normalize review items. - Convert each comment into: `id`, `summary`, `affected_files`, `requested_change`. 2. Validate each item against the actual code. - Read referenced files and surrounding context. - Classify item status: - `valid-defect`, - `valid-improvement`, - `not-applicable`, - `needs-clarification`. - Capture concrete evidence (paths, lines, behavior). 3. Build an implementation plan. - Include only validated and accepted items. - Order tasks by dependency and file overlap. - Define acceptance criteria and verification commands per task. 4. Delegate tasks to subagents. - Select implementation/testing subagent by task type. - Pass review IDs, evidence, constraints, and checks. 5. Reconcile final status. - For each original item, report `fixed`, `declined-with-rationale`, or `blocked`. ## Required Output 1. Validation matrix (`ID | Verdict | Severity | Evidence | Rationale`) 2. Ordered implementation plan with task owners 3. Subagent dispatch prompts 4. Final reconciliation ledger ## Rules - Do not implement before validation. - Keep explicit rationale for declined or out-of-scope items. - Prefer smallest safe fix that satisfies the concern. - If review items conflict, call it out and propose resolution options.