Plugin Claude Code
Give your AI agent an operating contract — twelve build principles, actively enforced.
Plugin Claude Code
Give your AI agent an operating contract — twelve build principles, actively enforced.
Plugin Claude Code
Give your AI agent an operating contract. Twelve build principles — the CEO/CTO division of labor, grill-me-before-you-build, reality-graded honesty — actively enforced through composable skills and a session-start hook. Stop your agent from being a yes-man.
Hook
Runs when a session starts, executing load-principles.sh. From evanmcgoogan/skin-in-the-game.
Skill Claude CodeCodex
A gate that self-reviews a plan or a diff against all twelve build principles before you build or before you call it done — the methodology's enforcement layer. Use this after drafting a plan and before executing it, before finalizing a change, or whenever the user says "check against the constitution / the…
Skill Claude CodeCodex
Before building or optimizing anything, question whether it should exist at all and try to delete it. Use this whenever you're handed a spec, feature list, requirement, metric, or "while you're at it" addition — especially before any optimization, new abstraction, or added config. Trigger whenever the user asks you to…
Skill Claude CodeCodex
Before building at the CEO line, interrogate intent with a tight set of sharp, first-principles multiple-choice questions — each option carrying pros, cons, and long-term implications, recommendation first. Use this whenever a substantial build hinges on a taste, strategy, or direction decision that is genuinely the…
Skill Claude CodeCodex
Establishes the operating contract between the human and the agent — the agent owns technical and reversible decisions outright and proceeds, escalating only genuine taste, strategy, or one-way-door calls (concisely, with options and a recommendation). Use this at the start of ANY non-trivial build, refactor, feature…
Skill Claude CodeCodex
Enforce honest, reality-graded reporting and kill the instinct to perform success. Use this whenever you're about to report results, claim something works or is "done," add a metric or success criterion, or summarize what you built. Trigger whenever you would otherwise say a task is complete, a test passes, or a…
Skill Claude CodeCodex
Read before you write and verify before you call it done. Use this before editing an unfamiliar codebase, implementing a feature, or changing existing code — understand the existing surface, patterns, and constraints first, then verify the result against real behavior. Trigger whenever the user asks you to implement…