llms.txt

# packforge

> AI Agent Instruction Hub — orchestrates context-aware instruction packs for Cursor, Claude Code, Codex, and other MCP-compatible agents.

## Overview

packforge analyzes your project context (stack, domain, phase, risk) and recommends the right instruction packs for your AI coding agent. It works as a Model Context Protocol (MCP) server.

## Architecture

- context-analyzer: analyzes project stack, domain, phase via package.json, file tree, and GitNexus graph
- orchestrator: matches and scores instruction packs against project context
- mcp-gateway: MCP stdio server exposing 8 tools
- shared-types: Zod schemas for packs, context, and activation

## MCP Tools

- start_project_from_spec: describe your project, get pack recommendations
- confirm_activation: activate packs, receive system prompts and bootstrap steps
- analyze_project_context: raw context analysis
- recommend_packs: raw pack matching
- activate_pack_set: activate specific packs
- get_activation_status: check current activation
- get_active_instructions: get active system prompts
- list_registry_packs: list all available packs

## Instruction Packs

17 packs across 5 categories: engineering, quality, ops, product, documentation.

Each pack contains: system prompt, tool permissions, activation signals, personality config, and execution policy.

## Key Concepts

- Pack: a YAML file defining an AI agent's behavior for a specific task domain
- Context: analyzed project metadata (stack signals, domain, phase, risk profile)
- Activation: the process of selecting and configuring packs for a project
- Bootstrap Steps: initial actions recommended when activating a pack set

## License

PolyForm Noncommercial 1.0.0