claude-codex-gemini-consensus

What problem does it solve?

The Claude-Codex-Gemini Consensus skill orchestrates a blinded, cross-model review workflow that ensures rigorous validation of plans, code changes, and analyses by three leading AI agents.

Core Features & Use Cases

• Blinded multi-model reviews to prevent anchoring and promote independent verification.
• SOP-based workflows covering Code Review, Analysis Planning/Execution, and Publication-ready Reporting.
• Consensus-building orchestration with round-based discussions and mandatory sign-off from all three models.
• Audit-ready logs and artifacts for traceability, reproducibility, and governance.

Quick Start

Use the skill to orchestrate a blinded consensus on a target project. Examples:

• Run independent reviews: ./scripts/review.sh codex path/to/target
• Start blinded discussion for a plan: ./scripts/blinded-consensus.sh plan analysis/plan.md --cd /path/to/project --rounds 3
• Synthesize final consensus and generate a report: ./scripts/consensus-review.sh codepath --blinded --rounds 3 This section uses plain text prompts and commands (no code blocks).

Note: The actual execution relies on locally installed Claude, Codex, and Gemini CLIs.

Quick Start

• Run a full blinded consensus on a codebase: ./scripts/blinded-consensus.sh code /path/to/project --rounds 3
• Validate an analysis plan: ./scripts/blinded-consensus.sh plan analysis/plan.md --cd /path/to/project --rounds 2
• Generate the final publishable report: ./scripts/consensus-review.sh code /path/to/project --blinded --rounds 3