Communitygithub.com

mbmd/AIPDLC

Injectable workflow packages that guide AI coding agents through professional software delivery — 11 packages from idea to test accountability.

AIPDLC 是什麼?

AIPDLC is a Claude Code agent skill that injectable workflow packages that guide AI coding agents through professional software delivery — 11 packages from idea to test accountability.

相容平台Claude Code~Codex CLICursor
npx skills add mbmd/AIPDLC

在你喜歡的 AI 中提問

開啟一個已預先載入此 Agent Skill 的新對話。

說明文件

AIPDLC — The AI-* PDLC Family

Injectable workflow packages that give AI coding agents structured expertise for professional software delivery.

What This Skill Provides

When installed, your AI agent gains access to 11 chained workflow packages covering the full Product Development Life Cycle — from raw idea to production-ready workspace with governance and test accountability baked in.

Each package makes your agent behave like a senior domain expert: a PMO advisor for initiation, a CTO/architect for design, a DevOps engineer for workspace setup, a compliance officer for governance, a QA lead for test strategy. No plugins, no APIs — just markdown-based procedural knowledge your agent reads and executes.

The Chain

╔══════════ PORTFOLIO LAYER · scope = MANY projects ══════════╗

   (optional)
    AI-ILC  ⇢  AI-PILC  ⇢  AI-PPM
    Decide it   Initiate it   Govern it (portfolio)

╚═══════════════════════╤═════════════════════════════════════╝
                         │
                      AI-FLO   Route it (package-to-package)
                         │
╔══════════ PROJECT LAYER · scope = ONE project ══════════════╗

    AI-POLC → AI-UXD → AI-ADLC → AI-DWG → AI-DLC v1 (build)
    Own it    Design UX  Design it  Prepare it

    AI-GCE + AI-TGE ── alongside (continuous quality) ──►
    Guard it  Test it

╚═════════════════════════════════════════════════════════════╝

Packages at a Glance

PackageWhat It DoesExpert Role
AI-ILCRaw idea → Approved Idea BriefProduct Manager
AI-PILCRequirement → Project Initiation PackagePMO Advisor
AI-PPMMultiple PIPs → Portfolio governancePortfolio Manager
AI-FLOPackage-to-package routingOrchestration Engine
AI-POLCPIP → Product Backlog PackageProduct Owner
AI-UXDPIP + PBP → UX Design PackageUX Designer
AI-ADLCRequirements → Architecture PackageCTO / Architect
AI-DWGAP + PBP + UXP → Ready-to-code workspaceDevOps Engineer
AI-GCEWorkspace → Compliance enforcementGovernance Officer
AI-TGEWorkspace → Test strategy & coverageQA Lead

How to Use

Quick start (single package)

Tell your AI agent:

Using AI-PILC, help me initiate this project from my requirements.
Using AI-ADLC, design the architecture for this system.
Using AI-DWG, generate a development workspace from this architecture.

Full chain

Start at AI-ILC or AI-PILC and let each package hand off to the next. Every stage has a human approval gate — the AI proposes, you decide.

Standalone or chained

Each package works independently. You can install one, a few, or all eleven. When chained, each detects its predecessor's output and enriches its own work. When standalone, it handles missing predecessors gracefully.

Key Principles

  • Human-in-the-loop — approval gates at every stage
  • Injectable — drop files into any workspace, no plugins
  • Professional quality — output reads as if written by a senior domain expert
  • Platform-agnostic — works with any AI assistant that reads workspace files
  • Chain-aware — packages hand off via state markers, but each works alone too
  • Adaptive depth — Minimal / Standard / Comprehensive tiers match project complexity

Supported Platforms

Works with 6+ AI coding platforms:

  • Kiro
  • Amazon Q Developer
  • Cursor
  • Claude Code
  • Cline
  • GitHub Copilot (partial — workspace instructions only)
  • Any AI assistant that reads workspace markdown files

Installation

After installing this skill, clone the package set into a temporary nested folder, run the installer against your workspace root, then delete the source. This keeps your workspace root clean — the installer only leaves .aiflc/pdlc/ (the package home) and pdlc-ws/ (your outputs) behind.

# 1. Clone into a temporary folder — NOT your workspace root
git clone https://github.com/mbmd/AIPDLC.git .aiflc-src/AIPDLC

# 2. Install into your workspace root (add --platform / --bundle to skip prompts)
./.aiflc-src/AIPDLC/installer/install.sh --target .

# 3. Remove the source — nothing unnecessary is left in your root
rm -rf .aiflc-src

On Windows (PowerShell):

git clone https://github.com/mbmd/AIPDLC.git .aiflc-src/AIPDLC
.\.aiflc-src\AIPDLC\installer\install.ps1 -TargetWorkspace .
Remove-Item -Recurse -Force .aiflc-src

The installer copies each package into .aiflc/pdlc/ and creates pdlc-ws/ for your outputs. For platform-specific detail, see each package's setup/INSTALL.md.

Learn More


Built on AIFLC by Mohammad Maheri — part of the AI Full Life Cycle project.

相關技能