Skip to content

PRD.md

Latest commit

 

History

History
160 lines (116 loc) · 5.28 KB

PRD.md

File metadata and controls

160 lines (116 loc) · 5.28 KB

AI Assist Kit - Product Requirements Document

Overview

AI Assist Kit is a Go library that provides a unified interface for managing configuration files across multiple AI coding assistants. It solves the fragmentation problem where developers must maintain separate configurations for each tool they use.

Problem Statement

The AI coding assistant ecosystem is fragmented:

  • Claude Code uses .mcp.json, .claude-plugin/plugin.json, CLAUDE.md
  • Gemini CLI uses gemini-extension.json, GEMINI.md, TOML commands
  • OpenAI Codex uses config.toml, AGENTS.md, Markdown prompts/skills
  • Cursor IDE uses mcp.json, .cursorrules, hooks.json
  • VS Code Copilot uses mcp.json, copilot-instructions.md
  • Windsurf uses mcp_config.json, hooks.json

Developers who use multiple tools must:

  1. Maintain duplicate configurations in different formats
  2. Manually sync changes across tools
  3. Learn each tool's specific syntax and conventions
  4. Risk configurations drifting out of sync

Solution

AI Assist Kit provides:

  1. Canonical Types: A single source of truth in JSON format
  2. Adapter Pattern: Tool-specific converters for each AI assistant
  3. Format Conversion: Automatic transformation between formats
  4. Schema Validation: JSON Schema for validation and IDE support

Target Users

  1. Individual Developers: Using multiple AI coding assistants
  2. Teams: Standardizing AI tool configurations across projects
  3. Tool Authors: Building plugins/extensions for multiple platforms
  4. Enterprise: Managing AI assistant configurations at scale

Core Features

v0.1.0 - Foundation (Completed)

Feature Description
MCP Configurations Manage MCP server configs across 8 tools
Hooks Lifecycle hook configurations for 3 tools
Context Project context (CONTEXT.json) to CLAUDE.md

v0.2.0 - Plugins & Commands

Feature Description
Plugin Manifests Canonical plugin definition with Claude/Gemini adapters
Commands/Prompts Command definitions with Claude/Gemini/Codex adapters
JSON Schemas Validation schemas for plugins and commands

v0.3.0 - Skills & Agents

Feature Description
Skills Skill definitions with Claude/Codex adapters
Agents Agent definitions with Claude adapter
Context Converters GEMINI.md and AGENTS.md generation

v0.4.0 - Settings & Rules

Feature Description
Settings Permission and sandbox configurations
Rules Coding guidelines (.cursorrules, copilot-instructions)

v1.0.0 - General Availability

Feature Description
CLI Tool aiassistkit generate/validate/convert commands
Full Documentation Complete API docs and examples
Stable API Backward-compatible API guarantees

Supported Tools

Tool MCP Hooks Context Plugins Commands Skills
Claude Code ✅ ✅ ✅ 🔜 🔜 🔜
Gemini CLI — — 🔜 🔜 🔜 —
OpenAI Codex ✅ — 🔜 — 🔜 🔜
Cursor IDE ✅ ✅ 🔜 — — —
VS Code Copilot ✅ — 🔜 — — —
Windsurf ✅ ✅ — — — —
Cline ✅ — — — — —
Roo Code ✅ — — — — —
AWS Kiro ✅ — — — — —

User Workflows

Define Once, Generate Everywhere

# Define canonical plugin spec
plugins/spec/plugin.json
plugins/spec/commands/release.json
plugins/spec/skills/version-analysis.json

# Generate for all platforms
aiassistkit generate plugins/spec/ --output plugins/

# Result:
plugins/claude/.claude-plugin/plugin.json
plugins/claude/commands/release.md
plugins/gemini/gemini-extension.json
plugins/gemini/commands/release.toml
plugins/codex/prompts/release.md

Convert Between Formats

# Convert Claude MCP config to VS Code format
aiassistkit convert .mcp.json --from claude --to vscode -o .vscode/mcp.json

# Convert Claude hooks to Cursor format
aiassistkit convert .claude/settings.json --from claude --to cursor -o .cursor/hooks.json

Validate Configurations

# Validate plugin spec against JSON Schema
aiassistkit validate plugins/spec/plugin.json --schema plugin

# Validate all specs in directory
aiassistkit validate plugins/spec/ --all

Success Metrics

  1. Adoption: Number of projects using aiassistkit
  2. Tool Coverage: Number of AI assistants supported
  3. Configuration Types: Number of config types supported
  4. Conversion Accuracy: 100% round-trip fidelity for conversions

Non-Goals

  • Runtime integration with AI assistants
  • AI assistant functionality (prompting, completion)
  • Cloud-based configuration management
  • Configuration encryption or secrets management

Dependencies

  • Go 1.23+
  • github.com/pelletier/go-toml/v2 (for Codex TOML support)

Related Projects