Skip to content

Execution Targets

VisionSpec generates specifications. Execution targets consume those specifications and turn them into working software.

What Are Execution Targets?

An execution target is an AI coding agent system that takes your reconciled spec.md and orchestrates the actual implementation. VisionSpec supports multiple targets, each with different strengths:

┌─────────────────────────────────────────────────────────────────┐
│                        VISIONSPEC                                │
│                                                                  │
│  MRD → Press → FAQ → PRD → TRD → TPD → IRD                      │
│                        ↓                                         │
│                    spec.md                                       │
│                        ↓                                         │
│              visionspec export <target>                          │
└─────────────────────────────────────────────────────────────────┘
                              ↓
         ┌────────────────────┼────────────────────┐
         ↓                    ↓                    ↓
    ┌─────────┐         ┌─────────┐         ┌─────────┐
    │ AI-DLC  │         │ SpecKit │         │   GSD   │
    │ (AWS)   │         │(GitHub) │         │         │
    └─────────┘         └─────────┘         └─────────┘
         ↓                    ↓                    ↓
    ┌─────────┐         ┌─────────┐         ┌─────────┐
    │GasTown  │         │GasCity  │         │OpenSpec │
    │         │         │         │         │(future) │
    └─────────┘         └─────────┘         └─────────┘

Available Targets

Target Best For Execution Model
AWS AI-DLC Enterprise workflows, multi-phase development Three-phase lifecycle with approval gates
GitHub SpecKit GitHub-native development, PR workflows Sequential task execution
GSD Fast iteration, parallel execution Wave-based parallel tasks
GasTown Complex multi-agent orchestration DAG-based bead execution
GasCity Role-based agent orchestration Agent-order architecture

The Export Workflow

Step 1: Generate Specifications

Use VisionSpec's Working Backwards methodology to create your specs:

# Initialize project
visionspec init my-project --profile enterprise

# Author source specs (human-written)
visionspec create mrd -p my-project
visionspec create prd -p my-project
visionspec create uxd -p my-project

# Synthesize derived specs (LLM-generated)
visionspec synthesize press -p my-project
visionspec synthesize faq -p my-project
visionspec synthesize trd -p my-project

# Evaluate and approve
visionspec eval all -p my-project
visionspec approve all -p my-project

Step 2: Reconcile

Generate the unified execution specification:

visionspec reconcile -p my-project
# Output: docs/specs/my-project/spec.md

Step 3: Export to Target

Transform spec.md into the target's format:

# Export to your chosen target
visionspec export aidlc -p my-project
visionspec export speckit -p my-project
visionspec export gsd -p my-project
visionspec export gastown -p my-project
visionspec export gascity -p my-project

Step 4: Execute

Use the target system to implement the specification. Each target has its own trigger pattern:

# AWS AI-DLC
"Using AI-DLC, implement the project based on the vision document."

# GitHub SpecKit
"Using SpecKit, execute the plan in .specify/"

# GSD
"Using GSD, execute PLAN.md"

Choosing a Target

See Choosing a Target for a detailed comparison to help you decide which execution target fits your workflow.

Key Concepts

Specification vs Execution

VisionSpec handles specification (what to build):

  • Market requirements (MRD)
  • Product vision (Press Release)
  • Functional requirements (PRD)
  • Technical architecture (TRD)
  • Test strategy (TPD)
  • Infrastructure needs (IRD)

Execution targets handle implementation (how to build it):

  • Code generation
  • Test creation
  • Build orchestration
  • Deployment

Framework vs Target

Frameworks (part of VisionSpec) define how you create specifications:

  • AWS Working Backwards
  • Google Design Docs
  • Stripe API-First
  • Lean Startup
  • Design Thinking
  • Jobs to be Done

Targets (external systems) define how specifications become code:

  • AWS AI-DLC
  • GitHub SpecKit
  • GSD
  • GasTown
  • GasCity

You can mix any framework with any target. For example:

  • Use Lean Startup framework → export to GSD for fast iteration
  • Use AWS Working Backwards → export to AWS AI-DLC for enterprise rigor
  • Use Stripe API-First → export to SpecKit for API-driven development

Next Steps