cautious-computing-machine--primitives-plugin

Loads the four Karpathy coding-with-AI rails (Think Before Coding, Simplicity First, Surgical Changes, Goal-Driven Execution) as one reference page. Use only when the user explicitly asks for "the rails", "the principles", "Karpathy principles", "Karpathy guidelines", or "how should I work in this repo".

Autonomously optimize any Claude Code skill by running it repeatedly, scoring outputs against binary evals, mutating the prompt, and keeping improvements. Based on Karpathy's autoresearch methodology. Use when: optimize this skill, improve this skill, run autoresearch on, make this skill better, self-improve skill, benchmark skill, eval my skill, run evals on. Outputs: an improved SKILL.md, a results log, and a changelog of every mutation tried.

This skill should be used to remove AI-generated artifacts and unnecessary code before committing. Integrates with desloppify CLI for quantitative scoring and directed fixes. Falls back to LLM-based pattern detection when desloppify is unavailable.

This skill should be used when building AI agents using prompt-native architecture where features are defined in prompts, not code. Use it when creating autonomous agents, designing MCP servers, implementing self-modifying systems, or adopting the "trust the agent's intelligence" philosophy.

Prevent feature creep when building software, apps, and AI-powered products. Use this skill when planning features, reviewing scope, building MVPs, managing backlogs, or when a user says "just one more feature." Helps developers and AI agents stay focused, ship faster, and avoid bloated products.

This skill should be used when solving hard questions, complex architectural problems, or debugging issues that benefit from GPT-5 Pro or GPT-5.1 thinking models with large file context. Use when standard Claude analysis needs deeper reasoning or extended context windows.

Verify which environment variable keys are present in .env files or shell environment without exposing their values. Use when you need to check env configuration or compare keys between .env files.

Audits system prompts and plugin configurations against Anthropic's Constitutional principal hierarchy to identify instructions that conflict with Claude's training, attempt to weaponize Claude against users, violate inalienable user protections, or exceed operator permission boundaries.

Create an isolated git worktree for feature development with automatic setup. Use when starting work on a new feature branch to get a clean, fully-configured workspace without polluting your main checkout.

Set up repositories optimized for AI agent collaboration and autonomous coding. Use when starting a new project, making repo "agent-ready" or "AI-friendly", setting up CLAUDE.md/AGENTS.md, configuring structured logging, adding pre-commit hooks, or organizing code for agent navigation.

Generate reasoning-based system prompts from product context. Takes product information as input, outputs a Constitution-aligned system prompt following the "onboarding document" structure.