# Harness Multi-Expert Review

> Run a structured multi-expert review with evaluation criteria, context isolation, three review rounds, convergence, and governance-ready outputs.

- Skill: `gabeujin/harness-multi-expert-review` (Agent Skill)
- Install (CLI): `npx skillmds@latest add gabeujin/harness-multi-expert-review`
- Raw SKILL.md: https://api.skillmd.com/api/skills/gabeujin/harness-multi-expert-review/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: Gabeujin (https://skillmd.com/u/gabeujin)
- Updated: 2026-09-10
- Page: https://skillmd.com/skills/gabeujin/harness-multi-expert-review

---


# Harness Multi-Expert Review

Use this skill when a task needs multiple expert lenses before the goal is frozen or before the implementation plan is finalized.

## Review Model

### Round 1: Independent Expert Review

- generate compact evaluation criteria from governance artifacts, code conventions, and the active task goal
- delegate only the relevant file set to each expert lens
- collect evidence-backed findings only

### Round 2: Cross-Review Synthesis

- merge duplicates
- resolve conflicting recommendations
- identify critical, important, and optional follow-up items

### Round 3: Final Improvement Plan

- confirm the reviewed goal and boundaries
- sequence changes by dependency and risk
- define the final plan that implementation must follow

## Suggested Expert Lenses

- architecture
- frontend
- backend
- security
- data
- platform and operations
- UX and accessibility
- product impact

## Output Expectations

- a concise evaluation criteria note
- one artifact per review round
- a final improvement plan that can be implemented without guesswork