sdd-execute-correctionslisted

# Execute SDD corrections If the caller limits execution to a batch, return after step 5 before starting another batch: `batch-completed` with pending IDs if the batch was approved, or `blocked` with evidence if the batch still has a failure or no task is eligible. When all tasks are complete, execute step 6 before returning. A batch does not end the review. 1. Fix one review by argument/context and require `codereview.md`. Inventory root and `done/` tasks by ID, findings, acceptance, dependencies, and files. If ambiguous, request a choice; if the plan is missing, direct to `sdd-plan-corrections`. **Output:** exact review, with no duplicates or missing/circular dependencies. If all are complete, continue to joint validation. 2. Select eligible tasks. Delegate one task per subagent using this skill, with the exact path and instruction to execute **only step 3**, without delegating, reviewing, or moving files. The coordinator executes steps 1, 2, and 4 through 6. Without subagents, execute sequentially and keep review separate from the author. Parallelize only without collisions in files, contracts, configuration, and test resources. **Output:** batch with exclusive owners and confirmed correction authorization. 3. In the executor, read the stable report and task once per version; then retrieve relevant PRD/TechSpec sections, code, and skills. Trace the cause and implement within the limits. Apply TechSpec validation; in desktop C#/.NET, omit E2E even in legacy comma