Bullpen gives your AI coding agent ten senior-dev instincts: push back on bad ideas, run it before saying "done," attack its own code, ask before guessing, and more. Ten skills you call by name at lite/full/ultra intensity. Open source, MIT, installable in Claude Code in two lines.

Hey Product Hunt 👋 I kept hitting the same wall with AI coding agents: they over-build, invent APIs that don't exist, and say "Done ✅" on code they never ran. A senior dev wouldn't. Their value isn't speed — it's instinct: they push back, verify, and try to break their own code before it ships. So I bottled 10 of those instincts into Bullpen. You call them by name while coding: • skeptic – pushes back on the wrong idea • closer – won't claim done until it ran the code • attacker – red-teams its own code before shipping • fact-checker, interrogator, stop-digging, doorman, historian, chameleon, explainer Each is one file, MIT, with lite/full/ultra intensity. Inspired by Ponytail (one lazy senior dev) — this is the whole room. Would love your feedback on which instinct is missing. 🧢

Love the named skills concept, feels way more natural than prompting from scratch every time. One thing I'd love to see is a per-skill memory or history view so I can see how often "push-back" or "attack-its-own-code" actually triggered in a session and what it caught. Would make it easier to trust the intensity settings and tune them to my codebase over time.

Two-line install was painless and the "attack its own code" skill actually caught a silly mistake I'd already missed. Feels less like a plugin and more like a sharper second pair of eyes.

Bullpen

Make your AI coding agent think like a senior dev

Make your AI coding agent think like a senior dev