AI-Driven Development skill
Over twenty skills for coding agents: MCP and hooks, plan mode, a bug-fix protocol, prompting and a bash safety hook
- 167 ★ on the repository
- Last commit
- MIT
Broad access
Why this level
- Configuration skills edit MCP configs, hooks and agent settings, which changes what the agent is allowed to do
- The macOS and Windows cleanup skills delete files, though they ask for separate confirmation first
- The hook is installed by a script that is downloaded and run through `curl` and `sh`
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/ai-driven-developmentDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- Configuration skills edit MCP configs, hooks and agent settings, which changes what the agent is allowed to do
- The macOS and Windows cleanup skills delete files, though they ask for separate confirmation first
- The hook is installed by a script that is downloaded and run through `curl` and `sh`
This is third-party code. Review the repository files before installing.
What it does
CodeAlive's repository gathers AI-assisted development practices into an umbrella set of 23 skills and one hook. The skills follow the Agent Skills standard, so they work in Claude Code, Codex CLI, OpenCode, Cursor, Gemini CLI, Antigravity and other agents. Each sits in its own folder and installs separately. The first group teaches an agent to configure itself: `mcp-management` installs MCP servers in ten or more agents and handles JSON, YAML and TOML differences, `hooks-management` and `settings-management` edit hooks and settings for Claude Code, Codex CLI and OpenCode, and `skills-management` trains skills in a loop with held-out validation. The second group keeps engineering discipline: `plan-mode` blocks changes without an approved plan, `bug-fix-protocol` treats every bug as two failures, the code and the test system, and `investigating-repository-history` reconstructs intent from `git blame` and pull request discussions. Separate skills cover prompting, OpenRouter provider ranking, Semantic Scholar search, converting pages to Markdown and macOS and Windows cleanup, while the `bash-guard` hook installs with its own script and parses shell commands before they run.
Who it is for. Developers who work with Claude Code, Codex or Cursor every day and want to configure the agent, introduce planning discipline and run commands safely.
Good fit when
- An MCP server has to go into several agents at once, and their configurations differ in format
- The agent edits without a plan and you want changes allowed only after approved steps
- A production bug should be closed in the code and in the tests that let it through
- Before a complex edit you want to understand why the code looks as it does, from `git` history and pull request discussions
- The agent's shell commands should be parsed before they run, with confirmation on dangerous ones
Not a fit when
- You need one small skill for a specific task: the umbrella set covers dozens of unrelated topics and you do not have to install everything
- The agent runs on Linux and you want disk cleanup and crash diagnostics: those skills are written for macOS and Windows only
- You need hard command bans: the `bash-guard` hook asks for confirmation by default because agents bypass bans by rephrasing
- You expect the App Store review skill in this set: it moved to the Vibe Stack repository
Example request
Use mcp-management to add the Context7 MCP server to Claude Code, Cursor and Codex at once without touching existing configuration entriesLimitations
The `bash-guard` hook is not part of the plugin set. It installs with its own `install-prebuilt.sh` script, and it is worth reading that script before running it in a shell. The `windows-qa-engineer` skill relies on Microsoft UFO and FastMCP. Every installed skill adds its description to the agent's context, so take the ones you need rather than all 23. Links to moved skills, such as `agents-consilium`, are listed separately in the README.
How to disable. Delete the folders you want from the agent's skills directory or run `/plugin uninstall ai-driven-development@ai-driven-development` in Claude Code. The `bash-guard` hook is removed from the agent's `settings.json`, where its install script registered it.
FAQ
How do I install the set in Claude Code?
Run `/plugin marketplace add CodeAlive-AI/ai-driven-development`, then `/plugin install ai-driven-development@ai-driven-development` and restart Claude Code. One skill installs with `npx skills add CodeAlive-AI/ai-driven-development --skill plan-mode`.
What does plan-mode do?
The skill makes the agent study context first, name ambiguities, contradictions and blockers, ask focused questions and write an evidence-based plan. The agent starts changing the project only after explicit approval and then validates the result.
How does bug-fix-protocol differ from an ordinary fix?
It treats a bug as two failures: a defect in the code and a gap in the tests. The protocol has eight steps, and the last one checks that the test system now covers the case.
What is bash-guard?
A Claude Code hook of type `PreToolUse:Bash`, written in Go. It parses commands into a syntax tree, spots dangerous ones such as `rm`, `git push -f`, `kubectl` and cloud API mutations, and asks for confirmation by default instead of blocking.
Which agents can use the set?
Skills are read by Claude Code, Codex CLI, OpenCode, Cursor, Gemini CLI, Antigravity, Copilot CLI and other agents that support Agent Skills. The hook is built for Claude Code specifically.
Discussion
No questions or reviews yet.
Related
Skills14 skills that find the gaps in a task before the code: interviews, plans, assumption tests and proof that a bug fix works
Claude Code Skills and Agents
Claude Code Skills & Agents
A Claude Code plugin bundling skills and sub-agents: deep research, image generation, GitHub work, code review and model switching
A set of 116 skills, 44 agents and safety hooks for Claude Code, Cursor, Codex, Gemini CLI and nine more editors
Sam's Agentic Coding
Sam's Agentic Coding Rules, Templates and Examples
Sam McLeod's personal kit for coding agents: about 60 skills, CLAUDE.md rules, commands, hooks, subagents and context tips
A set of Claude Code utilities: background skills, expert agents, slash commands and prompt templates for development
The author's personal set: shared agent rules and dozens of skills for GitHub, the browser, macOS, messaging and releases, plus a sync script
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Repository stars | Main difference |
|---|---|---|---|
| AI-Driven Development (this card) | no | 167 | |
| Agent Skills by Addy Osmani | no | 113,521 | 25 engineering skills along the development cycle from spec to release |
| Superpowers | no | 313,096 | a skill set that sets the agent's development process with TDD and subagents |
| Anthropic skills | no | 188,047 | Anthropic's official example skills for documents, design and the MCP builder |
README in short
The README announces a set of 23 skills and one hook for many agents and lists them by group: agent self-configuration, engineering practices, AI infrastructure, research and docs, system utilities. Each skill gets a short paragraph and a link. Four design principles and the repository layout with `skills/` and `hooks/` follow.