albertomarturelo

Context First Development — AI skill for Claude Code

AI community

CFD is a methodology for engineers who use AI coding agents (Claude Code, Gemini CLI, Codex, Cursor, Aider) on real repositories.

How to install Context First Development

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

What Context First Development does

CFD is a methodology for engineers who use AI coding agents (Claude Code, Gemini CLI, Codex, Cursor, Aider) on real repositories.

Alternatives in AI

Claude Skill Web Clone — Methodology for faithfully cloning any website (static / React / WebGL) — without copying AI-hallucinated code 969 ★
Oc Claw — A desktop pet that monitors your AI coding agents (OpenClaw, Claude Code, Codex, Cursor, Gemini CLI, opencode 352 ★
AI Toolkit — Professional-grade AI coding toolkit: 94 skills, 44 agents, multi-platform (Claude, Cursor, Windsurf, Copilot 167 ★

README

Context-First Development

[![Validate catalog](https://github.com/albertomarturelo/context-first-development/actions/workflows/validate-catalog.yml/badge.svg)](https://github.com/albertomarturelo/context-first-development/actions/workflows/validate-catalog.yml)

Your AI agent has amnesia at the start of every session. CFD is the cure.

**CFD is a methodology for engineers who use AI coding agents** (Claude Code, Gemini CLI, Codex, Cursor, Aider) **on real repositories.** It treats context as a first-class artifact: structured, versioned, and reviewed like code.

**The primary payoff is adherence, not tokens**: code that respects your conventions on the first attempt, decisions that don't get re-litigated, corrections that never repeat. Token savings are a welcome side effect — orientation drops to a few thousand tokens — but if you adopt CFD to save tokens you're optimizing the wrong variable. Modern agents don't naively scan whole repos; they search. What they *can't* search is the **why**: the rejected alternatives, the conventions, the in-flight state. That is what CFD persists.

**Evidence.** Measure the ceiling on your own project with [`scripts/measure-session-cost.sh`](scripts/measure-session-cost.sh): it compares a CFD-style `/session:start` (reading the `CLAUDE.md` + `CURRENT_STATUS.md` + decisions-index triad) against reading every file — an **upper bound**, since a real agent explores selectively. The honest per-session metrics (re-explanation rate, time-to-first-correct-action) are defined in the methodology essay's Metrics section; measure those on real sessions.

Real-world adopter you can click through: **[`sii`](case-studies/sii.md)** — a TypeScript monorepo (shared core

CLI + MCP server for Chile's tax authority) born with CFD on day 0: ADR-001 is the adoption decision, 21 ADRs landed in the first week, CLAUDE.md's Critical Rules cite their backing ADRs by number, and the packages ship on public npm under MIT. A team case study from Equ