AI Board Review workflow for multi-perspective solution design
Invoke this skill in Claude Code:
Add --skip-critique to bypass the critique loop
Three AI models (Claude, Gemini, Codex) independently analyze your requirements, then their best ideas are synthesized into a single consolidated solution. The optional critique loop challenges the result for over-engineering.
Analyzes requirements and proposes a primary solution with architecture, tech stack, and implementation approach.
Focuses on high-level reasoning, system integration, user flows, and architectural coherence.
Focuses on technical implementation feasibility, code structure, stack choices, and practical concerns.
Best components from all three are merged, conflicts resolved, producing a consolidated solution document.
After the initial solution, all three AI models switch roles from designer to critic. They evaluate for over-engineering, identify simpler alternatives, and flag red flags. The consensus drives a v2 that's leaner and more robust.
Rates complexity (1-10), proposes "boring technology" alternatives, identifies what to remove to cut 30%.
Estimates lines of code, audits each dependency, scores deployment complexity, proposes concrete simpler tools.
Challenges own assumptions, identifies invented requirements, rates confidence, finds the "90% solution that's 50% simpler".
{descriptor}.md - Refined requirements{descriptor}.claude.md - Initial analysis{descriptor}.gemini.md - Gemini design review{descriptor}.codex.md - Codex design review{descriptor}.final.md - Final solution (v1)index.html - Executive presentation{descriptor}.final.v1.md - Archived first synthesis{descriptor}.critique.gemini.md - Gemini critique{descriptor}.critique.codex.md - Codex critique{descriptor}.critique.claude.md - Claude self-critique{descriptor}.refactor.md - Refactoring plan
# Full workflow with critique (11 steps)
/createsolution Build a stock portfolio tracker
# Skip critique loop (7 steps, faster)
/createsolution --skip-critique Build a stock portfolio tracker
# Output locations:
# Markdown: ~/projects/ainotes/shared/createsolution/
# HTML: nginx.ai-servicers.com/{username}/createsolution/{descriptor}/