Skip to content

DerianAndre/aidd.md

Repository files navigation

aidd.md logo

aidd.md — AI-Driven Development

The open standard for AI-Driven Development. Multi-IDE, AI-agnostic agent coordination.

Version: 1.0.0 Last Updated: 2026-02-18 License: MIT Domain: https://aidd.md


What is AIDD?

AIDD is an open standard for structuring AI-assisted development. It provides a modular, composable framework of agent roles, rules, skills, workflows, templates, and specifications — all under .aidd/content/ in your project — that any AI system can consume, enabling consistent, evidence-based development across IDEs and providers.

Core principles: Evidence-First engineering, First Principles thinking, BLUF-6 communication, Zero Trust verification, Lean Antifragility.

Key benefits:

  • Multi-IDE compatible (Claude Code, Cursor, Gemini, Warp)
  • AI-agnostic (works with Claude, Gemini, ChatGPT, local models)
  • Modular and composable (load only what you need)
  • Evidence-based (every decision traceable to data or principles)
  • Progressive disclosure (token-efficient context loading)
  • Self-improving (session memory, pattern detection, framework evolution)

How to Follow the Methodology

Read WORKFLOW.md for the full user guide — core concepts, conversation lifecycle, automatic vs manual operations, and tips for getting best results.

Quick Start

Contributors (clone the repo)

git clone https://github.com/DerianAndre/aidd.md.git
cd aidd.md
pnpm install
pnpm setup          # builds, detects IDEs, configures, verifies
# Open your IDE — MCPs auto-start

Adopters (use in your project)

Option A — CLI (recommended):

cd my-project
npx @aidd.md/cli init    # copies framework, sets up .aidd/, integrates IDEs
# Open your IDE — MCPs auto-start

Option B — Manual:

# Copy the AIDD framework into your project
mkdir -p your-project/.aidd/content
cp -r aidd.md/content/ your-project/.aidd/content/
# Create thin AGENTS.md redirect at root (for Gemini compat)
echo "See [.aidd/content/agents/](.aidd/content/agents/) for agent definitions." > your-project/AGENTS.md

Activate the agent system

Use this prompt with any AI:

Read .aidd/content/agents/routing.md and assume the most appropriate role for my task.

IMPERATIVES:
1. Respect ALL rules in .aidd/content/rules/global.md
2. Load domain-specific rules as needed
3. Use skills from .aidd/content/skills/ directories
4. Follow workflows in .aidd/content/workflows/ for multi-step tasks

Confirm your role and loaded rules before proceeding.

Core Concepts

Agent System

11 specialized roles coordinated by the Master Orchestrator:

Agent Purpose Activation
Master Orchestrator Logic orchestration, context validation Entry-point for any task
System Architect C4 diagrams, ADRs, system design "architecture", "system design"
Contract Architect OpenAPI 3.1, API governance "API design", "OpenAPI"
Data Architect SQL schemas, data strategy "database", "schema"
Interface Artisan React components, WCAG, Tailwind "component", "UI"
Experience Engineer Frontend architecture, performance "state management", "optimize"
Design Architect Design systems, tokens "design system", "tokens"
Security Architect OWASP audits, defense-in-depth "security", "vulnerabilities"
Quality Engineer Test generation, quality strategy "write tests", "coverage"
Platform Engineer CI/CD, Docker, monitoring "deploy", "infrastructure"
Knowledge Architect Documentation sync "update docs", "document"
i18n Specialist Internationalization "i18n", "translate"

Rules (13 files)

Immutable constraints — prescriptive, MUST follow. Located in content/rules/:

Rule Scope
global.md Evidence-First, BLUF-6, heuristics, immutability constraints
orchestrator.md Task decomposition, strategic mapping, TKB integration
backend.md Hexagonal architecture, API design, security patterns
frontend.md WCAG 2.1 AA, React patterns, Tailwind best practices
testing.md Test pyramid (70/20/10), AAA pattern, coverage targets
security.md OWASP Top 10 (2025), secret management, CI/CD security
interfaces.md Zero-fidelity contracts between agents
code-style.md Naming conventions, TypeScript strict, import patterns
git-workflow.md Commit format, branch strategy, spec-first flow
documentation.md Normalized patterns, folder structure, ADR format
performance.md Performance budgets, optimization patterns
deliverables.md Output format standards and deliverable requirements

Specifications (13 files)

Formal reference documents — descriptive, detailed rationale. Located in content/specs/:

Specs Content
aidd-lifecycle.md 6-phase AI-Driven Development lifecycle
bluf-6.md 6-part communication protocol
heuristics.md 10 decision heuristics with examples
version-protocol.md 4-step version verification
memory-layer.md Project memory integration (decisions/mistakes/conventions)
supported-agents.md Agent registry and role definitions
model-matrix.md Multi-provider model routing matrix
prompt-evolution.md Prompt evolution and optimization patterns
orchestrator.md Multi-agent workflow orchestration standards
fast-track.md Per-phase workflow skip for trivial tasks
implementation-summary.md Structured completion report for SHIP phase
ltum-aidd.md Low Token Usage Mode protocol
routing.md Task classification and agent routing spec

Skills (11 directories)

Specialized capabilities with optional validation scripts. Located in content/skills/. Each skill has a SKILL.md with YAML frontmatter defining name, description, and recommended model tier.

Workflows (11 files)

Multi-step procedures located in content/workflows/: orchestrator, analyze, audit, design, docs, feature-branch, full-stack-feature, product, review, technology-selection, test. The orchestrator is THE single entry point for all tasks.

Templates (5 files)

Standalone protocol templates located in content/templates/. Domain-specific content has been absorbed into content/rules/, content/skills/, and content/workflows/. See content/templates/routing.md for the full routing table.

Template Purpose
brainstorming.md Diverge → Analyze → Converge → Document protocol
research.md Multi-source investigation and synthesis
penetration-testing.md Authorized offensive security testing
migration.md Version upgrades and breaking change management
routing.md Task-to-template/rule/skill/workflow routing (SSOT)

Knowledge Base (106 entries)

Quantified technology metrics across 13 categories located in content/knowledge/: runtimes, frontend, backend, data, testing, infrastructure, security, tooling, patterns. The Master Orchestrator queries the TKB for evidence-based technology recommendations.


AIDD Lifecycle

AI-Driven Development — the recommended development workflow:

PLANNING (Tier 1)              EXECUTION (Tier 2)         CLOSURE (Tier 3)
──────────────────             ──────────────────         ────────────────
1. UNDERSTAND (context+reqs)   4. BUILD (implement)       6. SHIP (commit+PR)
2. PLAN (atomic tasks)         5. VERIFY (test+check)
3. SPEC (persist spec)

See content/specs/aidd-lifecycle.md for full details and WORKFLOW.md for the user guide.


MCP Ecosystem

The AIDD MCP engine exposes 82 tools across 5 packages for full session tracking, memory persistence, validation, and self-improvement:

packages/
  shared/           @aidd.md/mcp-shared    (types, utils, server factory)
  cli/              @aidd.md/cli           (CLI — framework management)

mcps/
  mcp-aidd-engine/  @aidd.md/mcp-engine    (engine — all modules in one process)
  mcp-aidd-core/    @aidd.md/mcp-core      (brain — guidance, routing, knowledge)
  mcp-aidd-memory/  @aidd.md/mcp-memory    (memory — sessions, evolution, analytics)
  mcp-aidd-tools/   @aidd.md/mcp-tools     (hands — validation, enforcement, CI)
Category Tools Purpose
Project + Guidance 6 Bootstrap, task routing, heuristic analysis
Knowledge + Agents 5 TKB queries, tech compatibility, agent skills
Model Routing 3 Model matrix, tier routing, matrix health
Sessions + Memory 16 Session lifecycle, observations, permanent memory, edits, integrity
Lifecycle + Evolution 12 Phase tracking, self-improvement (analyze, approve, reject, delete)
Analytics + Patterns 9 Model performance, output quality, fingerprinting
Validation + CI 19 Code quality, compliance, commit generation
Drafts + Artifacts 5 Content drafts (create/list/approve/reject), workflow documents
Context + Diagnostics 7 Context optimization, scaffolding, health trends, system diagnostics

See mcps/README.md for full architecture and mcps/PLAN.md for implementation roadmap.


Directory Structure

aidd.md/
├── AGENTS.md                    # Thin redirect to content/agents/ (Gemini compat)
├── CLAUDE.md                    # Claude Code conversation lifecycle protocol
├── WORKFLOW.md                  # User-facing methodology guide
├── README.md                    # This file
├── LICENSE                      # MIT
├── CONTRIBUTING.md              # Contribution guidelines
├── package.json                 # Monorepo root
├── pnpm-workspace.yaml          # Workspace config
│
├── content/                     # Framework content (SSOT)
│   ├── agents/                  # Agent definitions + routing
│   ├── specs/                   # Formal specifications (13)
│   ├── rules/                   # Immutable constraints (13)
│   ├── skills/                  # Agent capabilities (11)
│   ├── workflows/               # Multi-step procedures (11)
│   ├── knowledge/               # Technology Knowledge Base (106 entries)
│   └── templates/               # Protocol templates (5) with routing
│
├── adapters/                    # IDE/AI integration guides
│   ├── README.md                # Adapter comparison and overview
│   ├── claude/                  # Claude Code (full: context + hooks + MCP)
│   ├── cursor/                  # Cursor (context + MCP)
│   ├── gemini/                  # Gemini/Antigravity (context + MCP + API)
│   └── warp/                    # Warp (context only)
│
├── packages/                    # Shared libraries and CLI
│   ├── shared/                  # @aidd.md/mcp-shared
│   └── cli/                     # @aidd.md/cli
│
├── mcps/                        # MCP server packages
│   ├── mcp-aidd-engine/         # @aidd.md/mcp-engine (all-in-one)
│   ├── mcp-aidd-core/           # @aidd.md/mcp-core (guidance)
│   ├── mcp-aidd-memory/         # @aidd.md/mcp-memory (sessions)
│   └── mcp-aidd-tools/          # @aidd.md/mcp-tools (validation)
│
├── apps/                        # Applications
│   └── hub/                     # AIDD Hub (Tauri 2 + React 19)
│
├── examples/                    # Project setup examples
│   ├── minimal/                 # Minimal AIDD setup
│   └── full/                    # Full AIDD setup with memory
│
├── docs/                        # Documentation
├── scripts/                     # Build and setup scripts
└── .claude/                     # Claude Code configuration
    ├── settings.json            # Permissions, hooks, status line
    └── statusline.js            # Cross-platform status bar script

IDE Integration

AIDD integrates with AI-powered development tools through three layers: Context (file-based), Protocol (conversation lifecycle), and MCP (82 tools). See adapters/README.md for the full comparison.

pnpm setup auto-detects and configures all supported IDEs:

IDE Detection Integration Layers Adapter Guide
Claude Code ~/.claude/ exists Context + Protocol + MCP adapters/claude/
Cursor Always available Context + MCP adapters/cursor/
Gemini AGENTS.md exists Context + MCP + API adapters/gemini/
Warp Manual setup Context only adapters/warp/

Run pnpm mcp:doctor to verify IDE integration status.


Contributing

See CONTRIBUTING.md for guidelines on adding rules, skills, workflows, templates, and knowledge base entries.


License

MIT License — See LICENSE.


Acknowledgments

Releases

No releases published

Packages

 
 
 

Contributors