sdlc-portfolio-manager: A Local-First Project Management Tool for Agentic Workflows

This tool combines the Azure DevOps thinking model with Cursor's native design to provide reusable rules, skills, and validation systems for the AI-assisted software development life cycle (Agentic SDLC). It features local-first and single-user-first as core principles, enabling efficient collaboration between human developers and AI Agents through three pillars: project management, rule library, and discovery-validation process.

With the popularity of AI coding assistants like Cursor, the collaboration model between developers and AI Agents has undergone fundamental changes. However, most existing project management tools were designed in the 'human-exclusive' era and lack native support for AI Agent workflows. Thus, sdlc-portfolio-manager was born as a local-first, single-user-first project portfolio management tool, using Cursor as the orchestration layer to allow humans and AI Agents to share work item storage.

The design draws inspiration from the Azure DevOps Boards mental model and the Paperclip control plane's ergonomics. The goal is to create a local, lightweight, Cursor-native project management tool that balances professional structure with AI workflow flexibility. It is currently in the Pre-alpha stage and uses OpenSpec specification-driven development (Proposal → Design → Specification → Task → Implementation → Archiving).

1. Portfolio and Work Item Management: Two-level hierarchy (Portfolio→Project→Work Item), bound to code repositories with shared storage; collaboration via HITL issues and @ mentions, with a unified inbox summarizing to-dos.
2. Rule and Skill Library Management: Centralized management of Cursor rules, skills, automation, and other assets, treating Agent behaviors as reusable, versioned assets that support cross-project reuse and auditing.
3. Discovery and Validation Process: The discovery phase turns unstructured ideas into user story drafts; the validation phase uses four gates (quality, security, etc.) and requires approval from isolated validators (skipping is allowed with notes).

Tech Stack: Monorepo structure, including Next.js+SQLite frontend, Cursor Agent call CLI, seed template library; database uses better-sqlite3 + Drizzle ORM, local-first with no SaaS dependencies. UI Principles: 8 principles including default quiet, single specification surface, progressive disclosure, keyboard-first, etc. Architecture Principles: Local-first, single-user first, Cursor-native non-exclusive, OpenSpec-driven, independent deliverable.

Quick Start: Use pnpm to manage dependencies with commands: pnpm install → pnpm rebuild better-sqlite3 → pnpm dev (visit http://localhost:3737); the first request automatically migrates and seeds data, and the storage path can be overridden via SDLC_DATA_DIR. Testing with vitest: pnpm test / pnpm test:watch. Development Status:

Section	Status
1. Monorepo + Toolchain	✅ Completed
2. Database Layer	✅ Partially Completed (basic tables established)
3. Portfolio + Project API	✅ CRUD + Seed + Health Check
4-21	Tracked in openspec/changes/initial-portfolio-manager/tasks.md

Core Value: Achieve human-machine shared work status, assetize Agent capabilities, data sovereignty autonomy, and specification-driven traceability. For teams exploring AI-assisted development, it is a reference implementation that combines the professionalism of traditional project management with AI flexibility, providing direction for building a true human-machine collaboration environment.