Mathews-Tom/armory

What it solves

armory provides a curated collection of production-grade skills, agents, and workflows for AI coding agents, specifically targeting Claude Code and Claude.ai. It solves the problem of inconsistent or "magic" AI outputs by providing battle-tested, opinionated automation units that define precise inputs, outputs, and failure modes for complex development tasks.

How it works

The project organizes its capabilities into several categories:

• Agents (Orchestrators & Analyzers): High-level agents (like team-lead or project-architect) that can decompose requests, delegate tasks to specialized agents, and synthesize results.
• Skills: Self-contained prompt or automation units for specific domains, such as GitHub operations, GPU optimization, systematic debugging, and architecture reviews.
• Planning Lifecycle: A structured sequence of packages (from decision-map to stacked-prs) that guides a feature from unresolved design to merged implementation.
• Runtime Model-Fit Matrix: A system for validating that the client and model (e.g., Claude Opus) are correctly configured for the required effort levels and token tracking.

Who it’s for

Developers who use AI coding agents as a serious, professional part of their software development workflow and require repeatable, high-quality results rather than simple demos.

Highlights

• Comprehensive Agent Ecosystem: Includes specialized agents for everything from security reviews and codebase auditing to business idea validation and research analysis.
• Structured Planning: A defined lifecycle for moving from unknown destinations to known designs and implementation slices.
• Surgical Tooling: Skills for advanced tasks like managing stacked pull requests, converting MCP servers to skills to save context window tokens, and automating Google NotebookLM.
• Quality Gates: Built-in review skills (e.g., pre-landing-review, architecture-reviewer) that apply severity-ranked findings to block or advise on code landing.