How does the Resolve stage work?

The Resolve stage requires streaming a minimum of 10 atomic 'micro-thoughts' to capture raw reasoning, constraints, and contradictions before a decision is finalized.

What are the stakes levels in FORGE?

Decisions are triaged into Low, Medium, High, or Critical levels based on their 'blast radius', security implications, and how easily they can be reverted.

Why is the Extract phase important?

The Extract phase (review_outcome) evaluates the actual results of a decision to refine future patterns and improve the engine's calibration over time.

How does this skill handle multi-agent environments?

The protocol uses strict scoping via agent_id and decision_id to prevent thought mixups and ensure isolation when multiple agents share an MCP connection.

What is the FORGE protocol?

FORGE stands for Fetch, Orient, Resolve, Go, and Extract—a systematic loop used within Claude Code to ensure high-quality, documented development decisions.

FORGE Decision Protocol

Name: FORGE Decision Protocol
Author: tfatykhov

bytfatykhov

0•

Productivity & Workflow

Implements a structured cognitive decision loop to track, validate, and optimize organizational judgment during development tasks.

FORGE is a specialized cognitive protocol designed for the Cognition Engine to formalize decision-making in software development. By following a five-stage loop—Fetch, Orient, Resolve, Go, and Extract—this skill ensures that every non-trivial code modification is backed by atomic reasoning and cross-referenced against historical patterns and guardrails. It utilizes a dedicated MCP server to maintain a compounding record of organizational judgment, helping developers avoid past mistakes, adhere to safety constraints, and distill successful patterns into reusable institutional knowledge.

Key Features

01Five-stage decision loop: Fetch, Orient, Resolve, Go, and Extract

02Pattern extraction to transform successful outcomes into institutional best practices

030 GitHub stars

04Atomic micro-thought tracking for transparent, step-by-step reasoning

05Automated guardrail validation and stakes-based triage for safety

06Historical context integration via session calibration and past decisions

Use Cases

01Safety-critical code modifications requiring rigorous reasoning and user confirmation

02Building a searchable repository of institutional knowledge and engineering logic

03Architectural decision-making for complex multi-service migrations

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add tfatykhov/cognition-engines-marketplace forge-protocol

For use in Claude.ai and ChatGPT

Download Skill