Skip to content
NEWMarketplace and your team: sell what you build, and get plans and changes that are tested before you see them. See what is new

Library / AI Agents & Assistants

AI Agents & Assistants

Stress-test plans and drafts with Gemini and Claude for a second opinion

Runs when a plan or draft is submitted through a form or from another workflow. Gemini argues against it and proposes a Plan B, and Claude recommends which option to pursue.

What it does

## Stress-test decisions with Gemini and Claude Submit a plan, decision or draft and get an honest second opinion before you commit. The two models come from different providers on purpose: they make different mistakes. ### How it works 1. Input comes from the form (logged-in n8n users only) or from another workflow. 2. *Normalize and Quote Input* wraps your text in quoted blocks, so instructions inside it are treated as content (prompt-injection guard). 3. Gemini plays devil's advocate and proposes a Plan B. 4. Claude weighs Plan A against Plan B and recommends A, B or a mix, with next steps and open questions. Both models answer in a fixed schema. 5. If Gemini fails, Claude takes over both roles. If the review can't be completed, the report says INCOMPLETE and gives no recommendation. ### Setup 1. Add a Google Gemini credential to *Gemini Devil's Advocate Model* and an Anthropic credential to *Claude Synthesis Model*. 2. Set your error workflow in the workflow settings. 3. Activate and open the form URL of *When Plan Submitted*. ### Customization - Swap either model sub-node for another provider. Keep the two different. - Call it from an AI agent or another workflow with plan, goal and context. It returns JSON. Your input is sent to both model providers. Don't submit personal or confidential data unless your contracts with them allow it.

The steps

  1. When Plan Submitted
  2. When Triggered by Workflow
  3. Normalize and Quote Input
  4. Conduct Devil's Advocate Review
  5. Gemini Devil's Advocate Model
  6. Parse Plan B Output
  7. Create Synthesis Prompt
  8. Evaluate and Recommend Plans
  9. Claude Synthesis Model
  10. Parse Recommendation Output
  11. Verify and Format Final Result
  12. Check Form Source
  13. Display Result in Form
  14. Return Result to Workflow Caller

Similar automations