Research First Builder
Make your coding agent justify architecture with real sources before it builds.
Install
npx skills@1.7.0 add IT-Shelter-Labs/research-first-builder --skill research-first-build --agent codex --copyclaude mcp add research-first-builder -- npx skills@1.7.0 add IT-Shelter-Labs/research-first-builder --skill research-first-build --agent codex --copy{
"mcpServers": {
"research-first-builder": {
"command": "npx",
"args": [
"skills@1.7.0",
"add",
"IT-Shelter-Labs/research-first-builder",
"--skill",
"research-first-build",
"--agent",
"codex",
"--copy"
]
}
}
}Install command taken from the project's README or registry entry. Always review code before running it.
README excerpt
Read full README on GitHub →Research First Builder
Make your coding agent justify architecture with real sources before it builds.
Don't build from vibes. Build from evidence.
Research First Builder is a portable agent skill by IT Shelter.
It turns real software references into ADOPT / REJECT / DEFER decisions, a minimal plan and linked verification.
The repository is research-first-builder; invoke the skill as research-first-build. rfb.py is its offline helper.
Preview: release and validation status. Русский · Quick start · Examples
Ask your agent
Codex:
$research-first-build Research and build a webhook inbox for one local process.
Keep dependencies minimal. Inspect relevant real projects first, explain what
you reject, and verify the implemented plan. Use docs/research-first/inbox.
Claude Code: start with /research-first-build and the same task. For any host, explicitly asking it to use
research-first-build is sufficient when its skill tool exposes that skill.
Install
From your target project, with Node 22.20+:
npx skills@1.7.0 add IT-Shelter-Labs/research-first-builder --skill research-first-build --agent codex --copy
This GitHub route was tested in a clean project without Git credentials. Installation does not establish native model compatibility; see the actual host status. For a folder copy, download the skill archive or use a checkout as described below.
Download/extract this package, or use your local checkout. Copy the whole
skills/research-first-build folder into your target project's skill directory:
| Local host | Destination inside your project |
|---|---|
| Codex / ChatGPT desktop local coding workspace | .agents/skills/research-first-build/ |
| Claude Code | .claude/skills/research-first-build/ |
| Cursor (beta) | .agents/skills/research-first-build/ |
| OpenCode (beta) | .agents/skills/research-first-build/ |
For desktop Codex, this is a folder copy in your coding project; the separate codex terminal command is not required.
Open that project and select/invoke the skill. If it does not appear, restart the coding session/application.
Ordinary chat without local file/research tools cannot execute this workflow.
For a local checkout, use:
npx skills@1.7.0 add "" --skill research-first-build --agent codex --copy
Change codex to claude-code, cursor or opencode. Review the installer's destination before accepting.
Manual copying needs no Node. Python 3.11+ is required for deterministic checks; the agent needs primary-source
access and local file tools. Without Python, a labelled manual review is possible, without deterministic readiness.
Source repository: IT-Shelter-Labs/research-first-builder. Installation details and host docs.
The animation is an edited artifact walkthrough of the included team-chat example, not an agent-session recording.
Why use it?
Research alone is cheap. Deciding which parts of a reference fit your task is harder.
Similar plugs
Makes your AI agent think like the laziest senior dev in the room. The best code is the code you never wrote.
Turn any codebase, with its docs, SQL schemas, configs, and PDFs, into a queryable knowledge graph. A /graphify skill for Claude Code, Cursor, Codex, and Gemini CLI: local…
Why use many token when few token do trick. Viral skill + proxy for coding agents that cuts 65% of tokens by talking like a caveman.
Production-grade engineering skills for AI coding agents.
The 1st Collaborative AI Design Agent Workspace. Local-first and open source. Create prototypes, websites, apps, dashboards, presentations, images, and videos with your team…