johngbl

Staff Engineer — AI skill for Claude Code

AI community

Staff-Level Engineering Meta-Skill for AI coding agents.

How to install Staff Engineer

This entry records only its repository, not the path inside it, so there is no exact command to give. Open johngbl/staff-engineer and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Staff Engineer does

Staff-Level Engineering Meta-Skill for AI coding agents. Evidence-based verified loops, strict TDD, anti-guessing discipline, and clean code.

Alternatives in AI

Anthropic Quickstarts — by Anthropic - Offers comprehensive development guides for three distinct AI-powered demo projects with standa 15.4k ★
Omniget — Udemy & Hotmart course downloader, YouTube downloader (yt-dlp GUI, 1,800+ sites) + desktop app for AI agents 13.9k ★
Karpathy — An agentic Machine Learning Engineer 1.4k ★

README

Staff Engineer — Pragmatic Verified Loops

![Version](https://img.shields.io/badge/version-1.0.0-blue) ![License](https://img.shields.io/badge/license-MIT-green) [![skills.sh](https://img.shields.io/badge/skills.sh-johngbl%2Fstaff--engineer-black)](https://skills.sh/johngbl/staff-engineer)

**XP with AI: Extreme Programming practices applied to AI-assisted development.** A single, self-contained SKILL.md that transforms coding agents into pragmatic Staff-level engineers. Evidence-based verified loops, strict TDD, anti-guessing discipline, and honest technical communication. Works with Claude Code, Cursor, Codex, Cline, OpenCode, and many more via [skills.sh](https://skills.sh). Simple by default — ceremony only when earned.

Why Does This Repository Exist?

AI coding agents are powerful but undisciplined by default. They hallucinate APIs, skip tests, accumulate monolithic files, agree with bad ideas to please the user, and declare "done" without mechanical proof. The **Staff Engineer** skill fixes this by embedding battle-tested engineering discipline directly into the agent's operating instructions.

The skill equips coding agents to:

Never guess when they can verify. Research the codebase, read official docs, or ask the user — in that order. Zero hallucinated APIs or packages.
Write complete, production-ready code. No // TODO, no placeholders, no lazy omissions. Every function, every error path, every edge case.
Enforce strict TDD with the Deletion Rule. Failing test first, then implementation. If production code is written before the test, the agent deletes it and restarts.
Validate everything mechanically. Linter (0 warnings), type-checker (0 errors), test suite (exit code 0). "I think it works" is not acceptable.
Be technically honest. If the user's request has a flaw, the agent says so and proposes the correct alternative. No sycophantic agreement.
Follow project conventions. Write as a developer who h