consensus-engine

Run parallel validators on identical journeys and synthesize a consensus report.

2|Updated Apr 8, 2026

Search Tags:#automation #validation #multi-agent #consensus #evidence #journeys #verdicts

Ready to run

npx skills add https://github.com/krzemienski/validationforge --skill consensus-engine

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: consensus-engine
Source: https://github.com/krzemienski/validationforge/tree/main/skills/consensus-engine
Command: npx skills add https://github.com/krzemienski/validationforge --skill consensus-engine

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Orchestrates a multi-agent consensus validation to increase confidence in high-stakes features.

Core Features & Use Cases

Spawn ≥2 parallel validators, each with an isolated evidence subdirectory
Receive identical journey lists and record independent verdicts
Synthesize per-journey verdicts into a single consensus verdict with a confidence tier
Trigger root-cause analysis when disagreements arise before finalizing the report

Quick Start

Launch 3 parallel validators against the same journeys and generate a unified consensus report.

Frequently Asked Questions about consensus-engine

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

FAQPage Schema

How do I use multi-agent consensus validation for high-stakes feature releases?▼

Multi-agent consensus validation spawns two or more parallel validators against identical journey lists to collect independent evidence. It synthesizes per-journey verdicts into a unified consensus report with a confidence tier, increasing release confidence for high-stakes features.

What is multi-agent consensus validation and when do I need it?▼

Multi-agent consensus validation is an orchestration process that runs multiple independent validators on the same journeys to increase confidence. You need it for regression reviews, pre-ship gates, and flaky-behavior detection when single-validator results are insufficient.

How do I run parallel validators to detect flaky behavior in test journeys?▼

Launch at least two parallel validators against the same journey list. Each validator records independent verdicts in an isolated evidence subdirectory, and the consensus engine synthesizes a unified report only after all validators complete their runs.

Can I trigger root-cause analysis when validators disagree on a journey verdict?▼

Yes, the consensus engine triggers root-cause analysis when validator disagreements arise. This happens after all validators complete and before finalizing the report, ensuring conflicting evidence is investigated before a final consensus verdict is synthesized.

What's the best way to orchestrate pre-ship validation gates with multiple validators?▼

The best approach is spawning at least two parallel validators that receive identical journey lists and record independent verdicts in isolated evidence subdirectories. The engine enforces the official contract and synthesizes a single consensus report with a confidence tier after completion.

Why does consensus validation isolate evidence subdirectories for each validator?▼

Consensus validation isolates evidence subdirectories to ensure each validator records independent verdicts without cross-contamination. This isolation is enforced by the engine's contract, allowing accurate synthesis of per-journey verdicts into a unified consensus report.