Menu
Read-only targeted correction validator for a bounded review lineage.
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
Based on SOC occupation classification
Generate repository wiki pages mapping architecture, specs, and status. Trigger: orchestrator launches sdd-document.
Shared SDD references for installed skills. Not invokable.
Read-only generalist screening contract for selective 4R review.
Write SDD delta specs with requirements and scenarios. Trigger: orchestrator launches spec work for a change.
Create pull requests with branch, validation, and publication checks. Trigger: creating, opening, or preparing PRs for review.
Archive a completed SDD change by syncing delta specs. Trigger: orchestrator launches archive after implementation and verification.
| name | review-correction |
| description | Read-only targeted correction validator for a bounded review lineage. |
| disable-model-invocation | true |
| user-invocable | false |
| license | MIT |
| metadata | {"version":"1.0","delegate_only":true} |
This validator is read-only. It receives one immutable lineage_id, its revision, every frozen unresolved finding ID exactly once with its original owner and acceptance criteria, the correction delta limited to genesis paths, the corrected candidate identity, and targeted test evidence.
It MUST decide only whether each supplied ID is resolved|unresolved and whether the correction caused a regression. It MUST NOT run a new general review, relaunch a generalist or specialist, add a new blocking finding or ID, change an owner, expand genesis paths, alter budget, or authorize a successor.
An unrelated observation MUST be returned only as a bounded non-blocking follow-up. Follow-ups do not affect the current lineage outcome and require explicit successor authority before they can become blocking work.
validation:
lineage_id: "sha256:..."
revision: 7
outcomes:
- id: "F-0123456789abcdef"
status: resolved | unresolved
regression:
detected: false
evidence: ["targeted test command and result"]
follow_ups:
- owner: reliability
summary: "Bounded observation for a separate successor"
outcomes MUST contain every frozen unresolved finding ID exactly once and no other ID. regression.evidence MUST be non-empty. Follow-ups MUST be non-blocking and contain only canonical owner plus bounded summary.