Plaesy: Constitution Kit

Version: 0.0.1

What is Plaesy Constitution Kit?

Plaesy Constitution Kit is an AI prompt framework built around a project constitution — a single governing source of truth (.plaesy/memory/constitution.md) that every phase (/assess, /implement, /optimize, /fix, /improve) reads before acting. Unlike spec-driven tools scoped to code alone, Plaesy drives all eight dimensions of a project from that constitution — technical, design, business, legal, marketing, financial, management, and product — through workflow automation from idea to production-ready deliverable, with anti-hallucination protocols enforcing evidence-backed output at each phase.

Why Plaesy Constitution Kit (vs. other spec-driven tools)

Most spec-driven development tools — including GitHub's own spec-kit — scope the "spec" to code: a technical plan for an AI coding agent to implement. Plaesy Constitution Kit treats the constitution, not the spec, as the source of truth, and applies it beyond code: an /assess:business or /assess:legal run is governed by the same constitution as /implement:technical. If your project needs more than code specified — pricing model, compliance posture, go-to-market messaging — those live in the same governed workflow, not a separate tool.

Agentic Agile frameworks like BMAD-METHOD solve an adjacent problem well: role-based agents (Analyst, PM, Architect) collaborating through a structured planning-to-build loop, with human-in-the-loop checkpoints. That agent-role discipline is one we share (see .plaesy/roles/). Where Plaesy differs is scope of governance: BMAD's loop is sized to a single software delivery effort, while Plaesy's constitution governs every dimension of a project — technical, design, business, legal, marketing, financial, management, product — under one evidence-backed source of truth, so a /loop or /improve pass on the business model is held to the same rigor as one on the codebase.

Quick Start

Install & Initialize

Plaesy Constitution Kit ships as a single cross-platform Go binary (plaesy).

Option 1: Automated Installation (Recommended) ⚡

Linux / macOS:


curl -sSL https://raw.githubusercontent.com/plaesy/spec-kit/main/install.sh | bash

Nushell (Windows, Linux, macOS):

nu -c (curl -fsSL https://raw.githubusercontent.com/plaesy/spec-kit/main/install.nu)

Windows PowerShell:


Invoke-WebRequest -Uri "https://raw.githubusercontent.com/plaesy/spec-kit/main/install.ps1" -OutFile install.ps1 -UseBasicParsing; .\install.ps1

The install scripts automatically:

✅ Detect your platform and architecture
✅ Download the release binary and verify its SHA-256 checksum (no checksum, or a mismatch, means nothing is installed)
✅ Run plaesy install, which copies the binary to ~/.local/bin (Linux/macOS) or %LOCALAPPDATA%\Plaesy\bin (Windows)
✅ Print the PATH line for your shell. They do not edit your shell profile.

To install a specific release instead of the latest, set PLAESY_VERSION, for example PLAESY_VERSION=v0.0.1.

To check a downloaded binary's build provenance with the GitHub CLI (releases built before provenance was added, such as v0.0.1, have no attestation):


gh attestation verify plaesy-linux-amd64 --repo plaesy/spec-kit

Option 2: Build from Source (requires Go 1.20+)


git clone https://github.com/plaesy/spec-kit.git
cd spec-kit/scripts
go build -o ../plaesy ./cmd/plaesy

# put the binary on PATH, then:
./plaesy init my-awesome-app

# Or initialize in current directory with specific AI
./plaesy init . --ai claude

Using any AI assistant with optimized prompts:


# Complete automation from idea to production-ready code
/start Build a privacy-first photo organizer that automatically groups images by event, location, and people

# AI will automatically execute ALL phases:
# 1. Project Analysis → 2. Technical Research → 3. Specification Generation
# 4. Implementation → 5. Quality Review → 6. Performance Optimization
# 7. Bug Resolution → 8. Completion Report with next steps

# Resume work anytime - executes remaining phases automatically
/continue

For Existing Projects: Documentation & Assessment

Before adding features to existing codebases, assess and document the project:


# Comprehensive project assessment
Chat: "/assess @tw ./legacy-application"

# Security-focused assessment
Chat: "/assess @security ./production-system"

# Architecture assessment for new feature planning
Chat: "/assess @architecture ./microservices-platform"

The /assess command generates complete project documentation, analysis, and AI-powered insights with autonomous implementation capabilities.

Commands

Every subcommand, matching docs/reference.md. Flags and examples for each one live there; this list is the index.


# Same commands on Linux, macOS, and Windows PowerShell
plaesy init                    # Scaffold a new Plaesy project
plaesy init <directory>        # Interactive AI selection in specified directory
plaesy init <directory> --ai <platform>    # Use specific AI platform in directory
plaesy reload                 # Refresh generated .plaesy/ files, keeping memory and context
plaesy analyze                 # Analyze current project structure and generate documentation
plaesy analyze                   # Build or query a knowledge graph of the project
plaesy eval                    # Run the search evaluation harness against a golden set
plaesy delegate                  # Send a bounded sub-task to a configured model provider
plaesy doctor                  # Check health of reach platforms, search index, and config
plaesy reach                   # Access external data sources (web, github, youtube, etc.)
plaesy validate                # Run every project-wide check
plaesy constitution validate   # Validate a generated constitution
plaesy memory validate         # Scan .plaesy/memory/ for external references
plaesy markdown validate       # Lint Markdown with the built-in Go linter
plaesy assumptions validate    # Audit ASSUMED tags left by autonomous runs
plaesy docx validate           # Validate a .docx file's OOXML structure
plaesy pptx validate           # Validate a .pptx file's OOXML structure
plaesy xlsx validate           # Validate a .xlsx file's OOXML structure
plaesy design validate         # Lint produced frontend output against the design rules
plaesy evals validate          # Run the instruction case files
plaesy config                  # Read and validate configs/platform.json
plaesy platforms               # AI platforms plaesy can target
plaesy tasks                   # Task lifecycle under .plaesy/tasks/
plaesy stack detect            # List instruction files for a target tech stack
plaesy trim                    # Token and context compression
plaesy images create           # Generate an image asset via a configured provider API
plaesy install                 # Install the plaesy binary to a well-known bin directory
plaesy loop                    # Run the autonomous loop driver (stop-hook: Stop hook for a running turn)
plaesy mcp                     # Serve the project's .plaesy/ folder over MCP (read-only)
plaesy uninstall               # Remove the installed plaesy binary
plaesy status                  # Check installation status and system information
plaesy clean                   # Clean current directory
plaesy clean <directory>       # Clean specified directory (default: current)
plaesy upgrade                 # Stub — prints "not yet implemented"
plaesy repair                  # Stub — prints "not yet implemented"

Popular AI Platforms

claude - Claude Code by Anthropic
cursor_ai - Cursor AI Assistant
github_copilot - GitHub Copilot
windsurf_ai - Windsurf AI

And 6+ other platforms available - see full list in documentation

AI Assistant Workflow Commands


/start <description>     # Begin new development workflow (generates project constitution first)
/assess [role] [path]    # Unified project assessment and deep research
/continue                # Resume and complete remaining phases
/implement               # Begin implementation phase with TDD (independent verifier pass before done)
/optimize                # Performance optimization and code refactoring
/improve                 # Best-practice gate for every active dimension -> routes defects/known gaps to /assess, /optimize, /fix
/loop                    # Autonomous assess -> fix -> verify -> repeat, no user input
/fix                     # Bug resolution and error recovery
/doc                     # Generate comprehensive project documentation
/save                    # Save current context and new knowledge
/create:images          # Generate image assets from text prompt or design spec (router for /create:{scope})

Documentation

Each top-level folder (scripts/, prompts/, templates/, instructions/, agents/, checklists/) holds the actual framework content consumed by the CLI and AI assistants. The matching folder under docs/ holds the human-readable guide for that content — start at docs/README.md for the full hub.

Component	Content	Documentation
Scripts	scripts/ — bash + PowerShell automation	docs/scripts/
Prompts	prompts/ — AI-optimized workflow prompts	docs/prompts/
Instructions	instructions/ — technology-specific guides	docs/instructions/
Templates	templates/ — project/document templates	docs/templates/
Agents	agents/ — AI role configurations	docs/agents/
Checklists	checklists/ — quality gate checklists	docs/checklists/
Testing	docs/testing/ — test strategy assets	docs/testing/

Version Information

Current Version: 0.0.1

Getting Version Information


# Check current version
cat VERSION

# Or use Plaesy CLI (when available)
plaesy --version