Claude Code Integration
Relevant source files
The following files were used as context for generating this wiki page:
Claude Code Integration within the Everything Claude Code (ECC) project represents a comprehensive framework designed to augment the capabilities of the Claude Code CLI tool through specialized agents, skills, and structured configuration. The system transforms the standard CLI into a production-ready AI coding harness by providing domain-specific intelligence for architecture, testing, security, and project management.
The integration is built upon an "Agent-First" philosophy, where complex tasks are delegated to a multi-tiered hierarchy of subagents and specialized workflows. This ensures that the AI possesses the necessary context, constraints, and tool access to perform high-fidelity code modifications, reviews, and architectural planning while adhering to project-specific conventions.
Sources: AGENTS.md:1-12, the-shortform-guide.md:21-25
System Architecture
The Claude Code integration utilizes a modular architecture composed of five primary layers: agents, skills, rules, hooks, and MCP servers. These components interact to provide a persistent, context-aware environment for autonomous and semi-autonomous development.
This diagram illustrates the hierarchical relationship between the main orchestrator and the various functional layers of the integration.
Core Component Breakdown
| Component | Purpose | Storage Location |
|---|---|---|
| Subagents | Specialized processes for delegated tasks with limited scopes. | ~/.claude/agents/ |
| Skills | Canonical workflow definitions and domain knowledge bundles. | ~/.claude/skills/ |
| Commands | Legacy slash-entry compatibility shims for tool-specific execution. | ~/.claude/commands/ |
| Rules | Mandatory guidelines (CLAUDE.md) that the AI must follow. | ~/.claude/rules/ |
| Hooks | Trigger-based automations firing on tool lifecycle events. | ~/.claude/settings.json |
Sources: the-shortform-guide.md:28-112, AGENTS.md:118-128
Context Management and Project Rules
The most critical component of the integration at the repository level is the CLAUDE.md file. This file acts as the primary source of truth for project-specific instructions, technology stacks, and coding conventions.
The CLAUDE.md Specification
The integration expects a CLAUDE.md file (typically under 100 lines) containing:
- Project Overview: Technology stack, versions, and port information.
- Quick Start: Copy-pasteable bootstrap and development commands.
- Architecture: A terminal-friendly directory tree with descriptions.
- Conventions: Naming styles, error handling patterns, and test file conventions.
Sources: agents/opensource-packager.md:46-99, skills/codebase-onboarding/references/summary-1.md:21-46
Project Onboarding Workflow
When entering an unfamiliar codebase, the integration employs a four-phase reconnaissance and mapping workflow to establish context.
The systematic onboarding process for generating project-specific context artifacts.
Sources: skills/codebase-onboarding/references/summary.md:15-60
Subagent Orchestration
The integration provides 67 specialized agents that the main orchestrator can delegate to. These agents are designed for specific domain expertise and utilize limited tool scopes to preserve context efficiency.
Priority Agents
| Agent | Trigger Condition | Primary Responsibility |
|---|---|---|
| Planner | Complex feature requests | Feature breakdown, dependency analysis, and phase planning. |
| Architect | Design decisions | System scalability and architectural pattern selection. |
| TDD Guide | Implementation tasks | Test-driven development (Red-Green-Refactor) execution. |
| Code Reviewer | Code modifications | Quality, maintainability, and pattern compliance checks. |
| Security Reviewer | Sensitive changes | Vulnerability detection (SQLi, XSS, Secret exposure). |
Sources: AGENTS.md:16-52
Parallel Execution Strategy
The system encourages parallelization by forking conversations or using multiple Claude instances in separate Git worktrees. This allows independent tasks (e.g., research vs. implementation) to proceed without overlapping context interference.
Sources: the-longform-guide.md:144-165
Implementation and Planning Workflows (PRP)
The Plan-Research-Postmortem (PRP) workflow is the standard for implementing new features. It emphasizes capturing all codebase patterns before writing code to ensure a single-pass implementation success.
Planning Phase Data Flow
The sequence of intelligence gathering during the PRP planning phase.
Sources: commands/prp-plan.md:16-118
Validation and Quality Gates
Integration with Claude Code includes mandatory validation steps for all implemented code:
- Static Analysis: Project-specific type checking (e.g.,
npm run typecheck). - Unit/Integration Tests: Running the detected test suite.
- Pattern Faithfulness: Verifying that new code is indistinguishable from existing codebase patterns (naming, logging, error handling).
Sources: commands/prp-plan.md:315-345, AGENTS.md:92-105
Event-Driven Automation (Hooks)
Hooks allow the integration to automatically execute actions based on tool usage, ensuring consistent quality without manual intervention.
Common Hook Implementations
- PreToolUse: Checking for
tmuxbefore long-running commands or preventing writes to unauthorized file types. - PostToolUse: Automatically running formatters (e.g., Prettier) or linters after code modifications.
- PreCompact: Saving the current state of work to a
.tmpfile before context compaction occurs to prevent loss of important session context.
Sources: the-shortform-guide.md:57-80, the-longform-guide.md:65-75
Conclusion
Claude Code Integration within the ECC project creates a robust ecosystem that transitions Claude from a general-purpose assistant to a highly specialized software engineer. By standardizing project rules in CLAUDE.md, utilizing a diverse array of specialized subagents, and enforcing rigorous planning workflows like PRP, the system ensures that AI-generated code is secure, maintainable, and contextually accurate to the target repository.