agent-instruction-forge

What problem does it solve?

This Skill helps you create effective, high-performance coding agent instruction files by extracting the real conventions and “why” behind how your team builds software, instead of relying on generic guidelines that agents cannot reliably follow.

Core Features & Use Cases

• Interactive rule extraction from your codebase: It reads existing instruction files and code to uncover implicit conventions, then guides you with targeted questions to surface non-obvious decisions and past failure modes.
• Greenfield and augment workflows: It supports starting from scratch (Greenfield) or improving existing instruction rules (Augment) with auditing, scoping fixes, and gap strengthening.
• Adversarial validation for reliability: It stress-tests the synthesized instruction file with isolated subagent challenges to catch omissions, weak rules, and contradictions before you deploy them.

Quick Start

Use agent-instruction-forge when you want Copilot/Claude Code/Cursor-style agents to follow your repository’s conventions by running the skill and answering the interactive questions after it discovers instruction infrastructure in your codebase.