plan-fix-reviewer

Reviews proposed plan fixes and outputs schema-conforming JSON approval decisions.

9|Updated Jun 10, 2026
One-click install
npx skills add https://github.com/eventbalancer/agent-quorum --skill plan-fix-reviewer-eventbalancer
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: plan-fix-reviewer
Source: https://github.com/eventbalancer/agent-quorum/tree/main/skills/plan-fix-reviewer
Command: npx skills add https://github.com/eventbalancer/agent-quorum --skill plan-fix-reviewer-eventbalancer

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? When an automated fixer patches stale or broken file:line references in an implementation plan, someone must verify the fix actually closes each finding without introducing new defects or weakening active invariants. This Skill performs that structured review and returns a machine-readable verdict. ## Core Features & Use Cases - Finding closure verification: Checks that every stale, ambiguous, or unresolved reference finding is genuinely fixed rather than masked or deleted. - Invariant assessment: Evaluates every active invariant occurrence exactly once with grounded evidence anchors, emitting satisfied, violated, not-applicable, or unresolved dispositions. - Schema-validated output: Produces JSON strictly conforming to review.schema.json with approval, coverage_complete, unresolved_occurrence_ids, invariant_assessments, and severity-tagged concerns. - Use Case: In the agent-quorum fix pass, after plan-fixer proposes edits to a canonical plan, this reviewer accepts, accepts with concerns, or rejects the proposal before it is applied. ## Quick Start Review the proposed fix in fix-proposal.md against the original plan and findings.json, then output the review verdict as JSON conforming to review.schema.json.

Frequently Asked Questions about plan-fix-reviewer

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I review a proposed plan fix automatically?

Supply the original plan, the proposed fix, and the findings list to the reviewer, which checks each finding for closure and returns a JSON verdict. The output conforms to review.schema.json with an approval field of accept, accept_with_concerns, or reject.

What output format does the plan fix review produce?

The review produces JSON only, with no markdown fences or prose. It contains approval, coverage_complete, unresolved_occurrence_ids, invariant_assessments with per-occurrence dispositions, and severity-tagged concerns, all validated against review.schema.json.

When is a proposed fix rejected during review?

A fix is rejected when at least one blocker or major concern exists, such as an unclosed finding, a new invalid file:line reference, lost plan sections, or heavy rewrites of unrelated content. Minor concerns alone yield accept_with_concerns.

How are invariant occurrences assessed in the review?

Every active invariant occurrence is assessed exactly once with a disposition of satisfied, violated, not-applicable, or unresolved. Conclusive dispositions require evidence copied exactly from the supplied deterministic candidate evidence anchors catalog.

What are the limitations of the fix reviewer?

The reviewer only assesses proposed reference changes and their side effects; it does not re-review the plan's general quality or propose new improvements. It also cannot approve an applied fix based on a prior proposal review.