MikeMoore1337/ai-development-toolkit

· Latest release · Skills catalog
Start without a terminal · I have an existing project · Local installation · Documentation
Already have a project? Start with the human-first guide. The coding agent contract is separate and is not required reading for a first request.
Toolkit integrations: ChatGPT plugin · Codex MCP · Windows, Linux and macOS setup · DeepSeek Chat · DeepSeek Harness
## Start with the task
AI Development Toolkit is a portable library of reusable instructions, specialized roles and QA workflows for AI-assisted development. It is not a standalone coding agent, hosting service or permission to access your project.
| You want to… | Start here | What to expect |
| Describe an idea without using a terminal | Beginner guide | Five copy-ready requests and honest limits for each AI client |
| Work on an already-open project | I already have a project | Read relevant guidance before proposing changes; no Toolkit installation required just to use the reference |
| Install and manage local skills | Safe installation | Profile selection, preview, explicit apply, status and rollback |
For a coding agent that can actually access your existing project, a useful first request is:
text
Use https://github.com/MikeMoore1337/ai-development-toolkit
to complete this task in my current project. Inspect first,
select only the relevant skills, implement and report what was actually tested.
Toolkit is a reference source, not an output directory for other users' code, logs or reports. An agent may only modify your project with appropriate tools and permissions. Merely linking a repository does not activate skills or prove execution.
## What this looks like in practice
These are illustrative workflows, not claimed real-world test runs.
| Starting request | Relevant path | Check before calling it done |
| “Design a distinctive fitness-club website.” | UX/UI Design Orchestra: compare concepts, critique, present actual previews | Owner reviews the visuals before UI implementation |
| “Find and fix a failing API test.” | QA Engineering Pack: reproduce, diagnose, make a focused change | Test results and the exact changed files are reported |
| “Help me prepare a release.” | Release quality and gates: risks, traceability, evidence | A human decides whether publication is authorized |
## What's inside?
Reusable skills for Frontend, Backend, Fullstack, QA, security, operations and UX/UI; specialist roles; and task-focused profiles for supported coding clients. Explore the current catalog rather than relying on counts embedded in graphics. The compatibility matrix distinguishes documented support from confirmed live execution.
Curated engineering and QA packs · How skills are evaluated · Make Feature · Clear Voice
Inventory counters (automatically checked in CI)
| Group | Inventory |
| Core skills | 24 |
| Frontend / Backend / Fullstack | 14 |
| QA Engineering Pack | 27 |
## Local MCP for coding agents
Fast start: Personal MCP - one-paste setup on Windows/Linux/macOS, plus ChatGPT/Codex local plugin marketplace. This reuses existing installers; public ChatGPT directory publication is a separate review.
Install Toolkit MCP on Windows, Linux or macOS - project-scoped Codex / Claude Code integration, read-only skill discovery and explicit config approval. ChatGPT plugin installation is separate.