Claude-Code on emsenn.net

ECC Hook Architecture Analysis

Sat, 04 Apr 2026 00:00:00 +0000

Everything Claude Code (ECC) is the most mature open-source Claude Code plugin, with 25+ hooks across 7 lifecycle events, a shared utility library, and a profile-gated execution framework. This analysis compares ECC’s hook architecture against the 561 harness to identify patterns worth adopting, hooks worth porting, and design decisions worth understanding.

¶Architecture Overview

¶ECC Hook Inventory

ECC registers hooks across every Claude Code lifecycle event:

Claude Code Development and Adoption

Mon, 30 Mar 2026 00:00:00 +0000

Claude Code originated as an internal engineering tool at Anthropic — built by the company’s own engineers to accelerate their work on Claude models. The trajectory from internal experiment to externally available product to billion-dollar revenue line was rapid, spanning roughly a year.

¶Timeline

February 2025 — Research preview. Claude Code launched publicly as a limited research preview alongside Claude 3.7 Sonnet. The initial form was a CLI tool (claude command, installed via npm) that operated as a terminal-based agent. It could read files, edit code, run shell commands, and manage git workflows, but required a terminal and developer familiarity to use.

Extension Architecture in Claude Code

Mon, 30 Mar 2026 00:00:00 +0000

Claude Code is extensible through four mechanisms. They operate at different levels of the system and serve different purposes. Understanding which mechanism to use for a given customization is the central design decision when configuring Claude Code for a project.

¶The four mechanisms

Mechanism	What it is	When it loads	What it affects	Deterministic?
CLAUDE.md	Markdown instruction file	Every session (root) or lazily (subdirectory)	Model reasoning and decisions	No — advisory
Skills	SKILL.md files invoked with `/name`	On demand (user or model invocation)	Model reasoning for a specific task	No — advisory
Hooks	Shell commands in settings.json	On tool-call lifecycle events	Runtime behavior before/after tool execution	Yes — mandatory
MCP servers	External processes exposing tools/resources	At session start (configured servers)	Available tool set and data sources	Yes — tool availability is deterministic; tool invocation is model-decided

¶The distinction that matters

The four mechanisms split along two axes:

CLAUDE.md

Sat, 07 Mar 2026 00:00:00 +0000

A CLAUDE.md file is a markdown document that Claude Code reads at the start of every conversation. It provides persistent, project-specific context that the agent cannot infer from code alone.

¶Load order

Claude Code loads CLAUDE.md files from four locations, additively:

Global (~/.claude/CLAUDE.md) — personal preferences, all projects
Project root (./CLAUDE.md) — committed to version control, shared
Subdirectory (path/to/dir/CLAUDE.md) — loaded lazily when Claude works in that directory
User-specific (.claude/CLAUDE.md) — personal overrides, gitignored

Ancestor files load at startup. Subdirectory files load lazily when the agent touches files in that directory. Sibling directories never cross-load.

Progressive Disclosure in Agent Instructions

Sat, 07 Mar 2026 00:00:00 +0000

Progressive disclosure is an information architecture principle applied to AI agent instruction files. The root instruction file (CLAUDE.md or AGENTS.md) contains only universally applicable rules. Task-specific, domain-specific, or tool-specific instructions live in subdirectory files, skill documents, or referenced specifications.

¶Why it matters

Every token in a root instruction file loads on every request. Long root files create three problems:

Budget pressure — instruction-following quality degrades as instruction count rises.
Relevance dilution — instructions irrelevant to the current task compete for attention with relevant ones.
Staleness risk — more content means more surface area for drift between instructions and actual project state.

¶The layering pattern

Layer	File	Loaded when	Contains
Root	`CLAUDE.md` (repo root)	Every session	Build commands, architectural constraints, what not to touch, pointers to key files
Subdirectory	`subdir/CLAUDE.md`	When working in that directory	Local conventions, tool-specific setup, domain rules
Skill	`skills/*/SKILL.md`	When skill is invoked	Detailed operational instructions for one task
Reference	Specification or guide files	When agent reads them on demand	Full style guides, formalization protocols, domain specs

¶In this repository

The emsenn.net repository applies progressive disclosure across two layers: