johnnichev

Nv:context — AI skill for Claude Code

AI community intermediate

npx skills add johnnichev/nv-context -g -y State-of-the-art context engineering for AI agents.

How to install Nv:context

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

What Nv:context does

**Context engineering for engineers who ship.**

Alternatives in AI

Codeburn — Free, local tool to track AI coding token usage and cost across 37 tools and agents (Claude Code, Cursor, Code 9.7k ★
Hindsight — State-of-the-art long-term memory for AI agents by Vectorize 6.7k ★
Gtm Engineer Skills — Claude Code skill for improving website AEO (AI Engine Optimization) and GEO (Generative Engine Optimization) 761 ★

README

nv:context

**Context engineering for engineers who ship.**

npx skills add johnnichev/nv-context -g -y

Auto-detects your tools, commands, and linters. Finds landmines with parallel agents. Generates configs, hooks, session management, and token budgets so every AI coding agent works at maximum effectiveness.

**Live experience:** [skills.nichevlabs.com](https://skills.nichevlabs.com) · production case studies, the research library, and the full synthesis.

Production Proof

Three real codebases. Same skill. Real numbers.

Repo	Before	After	What changed
selectools (Python SDK, 4,612 tests)	L3, 49/60	L5–L6, 58/60	CLAUDE.md 440 → 67 lines (-85%). Token budget cut 53%.
nichevlabs (multi-product SaaS, orchestrated by selectools)	L4, 17/60	L6, 49/60	SESSION.md 805 → 59 lines (-93%). Saved 15.8K tokens per session. 81 bugs surfaced.
sheriff (Python + TypeScript)	L4, 36/60	L5, 42/60	Incremental polish on an already-strong setup. Coverage gate added.

[Read the full case studies →](https://skills.nichevlabs.com/case-studies)

**Benchmark:** 100% pass rate on 3 test scenarios vs 45.8% baseline (+54.2% delta).

"Most agent failures are not model failures. They are context failures." Philipp Schmid, Google DeepMind

The Problem

Your AI agent is smart. But it keeps running the wrong test command, touching files it shouldn't, and forgetting your rules mid-session. You've tried writing a CLAUDE.md. Maybe you ran `/init`. It didn't help. Research proves why:

Finding	Source
Auto-generated configs reduce success by 3%	ETH Zurich
Experienced devs are 19% slower with bad AI context	METR Study
40%+ of AI project failures stem from poor context	IntuitionLabs
Using 40% of context window outperforms using 90%	Dex Horthy