FAQ

General

What is nv:context? A Claude Code skill that sets up state-of-the-art context engineering for any repository. It analyzes your codebase, interviews you about your workflow, and generates the complete infrastructure (config files, hooks, session management, token budgets) so AI coding agents work at maximum effectiveness.

What is context engineering? The systematic discipline of providing AI agents with the right information, in the right format, at the right time. It's the evolution beyond prompt engineering: instead of crafting clever prompts, you build the information ecosystem around your agents.

Why not just use /init? ETH Zurich research proved that auto-generated config files REDUCE agent success by 3% and increase costs by 20%+. nv:context interviews you first, analyzes your codebase for non-obvious patterns, and generates minimal, high-signal configs. Every line earns its place.

Which AI tools does this work with? The generated AGENTS.md works with 25+ tools including Claude Code, Cursor, GitHub Copilot, Windsurf, Aider, Gemini CLI, and more. Tool-specific configs (CLAUDE.md, .cursor/rules, etc.) are only generated for tools you actually use.

Setup

How long does setup take? ~2 minutes for the interview, ~30 seconds for analysis and generation. Under 3 minutes total.

Can I run it on an existing repo with configs? Yes. nv:context analyzes existing configs, scores them, and improves them rather than replacing wholesale. It will flag stale content, negative instructions, and missing critical sections.

Do I need to install anything besides Claude Code? No. nv:context is a Claude Code skill. Copy the SKILL.md file and you're done.

Technical

How many lines should my CLAUDE.md be? Under 200 lines per root file. Under 100 is ideal. Under 50 is elite. The research is clear: shorter files with higher signal density outperform longer files.

Why do negative instructions backfire? Research shows that "don't use X" increases the probability of the model using X. The instruction draws attention to X. "MUST use Y" directs attention to the correct alternative. Exception: "NEVER" as an absolute prohibition works fine.

What's the PostCompact hook? When Claude's context gets compressed (compaction), it can lose critical rules. The PostCompact hook automatically re-injects the most important context after compaction fires. It solves the #1 complaint: "the agent forgets my rules mid-session."

What's the hierarchy of leverage? A framework for prioritizing context engineering effort: verification > CLAUDE.md > hooks > skills > subagents > session management. Most people optimize from the bottom up. The best engineers start at the top.

What's the token budget report? An estimate of how many tokens your config files consume before the conversation even starts. If your configs eat 60%+ of the context window, there's not enough room for the actual task. nv:context warns you when this happens.

Philosophy

Why "for engineers who ship"? Context engineering matters most for people building real software with AI agents. Demo-quality prompting is easy. Production-quality context is hard. nv:context is built for the latter.

What's the METR study? A controlled study showing experienced developers are 19% slower with AI tools, despite feeling 24% faster. The 39-percentage-point perception gap means bad context actively hurts. Good context is what flips the equation.

Why is the research included in the repo? The 12 research logs (471KB digesting 200+ external sources) are the most comprehensive compilation on context engineering that exists. They're the foundation the skill is built on, and they're valuable as a standalone reference for anyone interested in the discipline. You can browse them at skills.nichevlabs.com/research or in the research/ folder.