# How the Planner Agent Activates for Feature Implementation in Claude Code

> Discover how the planner agent activates for feature implementation in Claude Code. Learn about rule-based orchestration and self-describing metadata for complex feature requests.

- Repository: [WorldFlowAI/everything-claude-code](https://github.com/WorldFlowAI/everything-claude-code)
- Tags: deep-dive
- Published: 2026-09-07

---

**The planner agent is automatically activated for feature implementation through rule-based orchestration and self-describing agent metadata that detects complex feature requests without requiring explicit user commands.**

When working with the `WorldFlowAI/everything-claude-code` repository, understanding how the **planner agent** springs into action helps you leverage its implementation-planning capabilities effectively. This specialized agent handles complex feature development and architectural changes through an intelligent dispatch system that reads your intent from natural language requests.

## Rule-Based Orchestration Triggers Automatic Activation

The orchestration layer in [`rules/agents.md`](https://github.com/WorldFlowAI/everything-claude-code/blob/main/rules/agents.md) defines explicit mappings between request types and agent assignments. For **complex feature requests** and **refactoring tasks**, the system mandates planner agent invocation without additional prompting.

The governing rule states:

> "Complex feature requests – Use **planner** agent"

This rule fires automatically when the dispatch logic parses your request. You don't need to manually specify the agent or add flags—the pattern matching happens behind the scenes.

## Self-Describing Agent Metadata Defines Activation Policies

Each agent in the repository carries a markdown header declaring its purpose and when it should activate. The planner agent's metadata in [`agents/planner.md`](https://github.com/WorldFlowAI/everything-claude-code/blob/main/agents/planner.md) explicitly instructs the system to use it **PROACTIVELY** for specific scenarios.

The header declares:

```markdown
---
name: planner
description: Expert planning specialist for complex features and refactoring.
Use PROACTIVELY when users request feature implementation, architectural changes, or complex refactoring.
Automatically activated for planning tasks.
---

```

This self-documentation serves dual purposes: it guides human developers and provides machine-readable instructions to the orchestration layer.

## Practical Activation Examples

### CLI Command Invocation

The standard way to trigger planner agent activation is through the `plan` command. The CLI parses natural language and routes to the planner when it detects feature-implementation patterns:

```bash
claude plan "Add OAuth2 login flow with token refresh"

```

The command handler in [`commands/plan.md`](https://github.com/WorldFlowAI/everything-claude-code/blob/main/commands/plan.md) forwards this request to the planner agent automatically.

### Internal Dispatch Logic

The orchestration system implements pattern matching at the code level. Simplified dispatch logic shows how request types map to agent launches:

```js
// Internal dispatch logic (simplified)
if (request.type === "feature" || request.type === "refactor") {
  // Rule "Complex feature requests – Use planner agent" fires here
  launchAgent('planner', request);
}

```

This ensures consistent behavior across all entry points—CLI, API, or internal calls.

### Resolution Flow

When multiple agents could potentially handle a request, the orchestration layer resolves conflicts using the explicit rules in [`rules/agents.md`](https://github.com/WorldFlowAI/everything-claude-code/blob/main/rules/agents.md). The planner agent takes precedence for:

- New feature development
- Architectural modifications
- Substantial refactoring operations

## Key Files Controlling Planner Agent Activation

| File | Purpose | Location |
|------|---------|----------|
| [`agents/planner.md`](https://github.com/WorldFlowAI/everything-claude-code/blob/main/agents/planner.md) | Agent definition with activation policy | [`agents/planner.md`](https://github.com/WorldFlowAI/everything-claude-code/blob/main/agents/planner.md) |
| [`rules/agents.md`](https://github.com/WorldFlowAI/everything-claude-code/blob/main/rules/agents.md) | Central rule set mapping requests to agents | [`rules/agents.md`](https://github.com/WorldFlowAI/everything-claude-code/blob/main/rules/agents.md) |
| [`commands/plan.md`](https://github.com/WorldFlowAI/everything-claude-code/blob/main/commands/plan.md) | CLI command implementation | [`commands/plan.md`](https://github.com/WorldFlowAI/everything-claude-code/blob/main/commands/plan.md) |

These files implement the complete activation pipeline. Modifying [`rules/agents.md`](https://github.com/WorldFlowAI/everything-claude-code/blob/main/rules/agents.md) changes when the planner engages; editing [`agents/planner.md`](https://github.com/WorldFlowAI/everything-claude-code/blob/main/agents/planner.md) updates its self-described capabilities.

## What Requests Trigger Planner Activation

The following patterns consistently activate the planner agent according to the source code:

- **Feature requests** – Any new functionality described by the user
- **Architectural changes** – Modifications to system structure or design
- **Complex refactoring** – Large-scale code reorganization

Simple requests that don't match these patterns route to other specialized agents instead.

## Summary

- **Rule-based orchestration** in [`rules/agents.md`](https://github.com/WorldFlowAI/everything-claude-code/blob/main/rules/agents.md) mandates planner agent use for complex features and refactoring without extra user input
- **Self-describing metadata** in [`agents/planner.md`](https://github.com/WorldFlowAI/everything-claude-code/blob/main/agents/planner.md) declares the planner should activate proactively for implementation planning
- The **`plan` CLI command** provides explicit access, but automatic pattern matching activates the agent from natural language alone
- Three core files—[`agents/planner.md`](https://github.com/WorldFlowAI/everything-claude-code/blob/main/agents/planner.md), [`rules/agents.md`](https://github.com/WorldFlowAI/everything-claude-code/blob/main/rules/agents.md), and [`commands/plan.md`](https://github.com/WorldFlowAI/everything-claude-code/blob/main/commands/plan.md)—implement the complete activation system

## Frequently Asked Questions

### Do I need to manually specify the planner agent when requesting a feature?

No. The orchestration layer automatically activates the planner agent when it detects feature implementation, architectural change, or complex refactoring patterns in your request. You can use the explicit `claude plan` command, but natural language descriptions also trigger automatic routing.

### What distinguishes "complex" feature requests that activate the planner from simple ones?

According to [`rules/agents.md`](https://github.com/WorldFlowAI/everything-claude-code/blob/main/rules/agents.md), the threshold centers on implementation planning requirements. Requests requiring architectural decisions, multi-step implementation strategies, or substantial code reorganization activate the planner. Simple, single-file modifications typically route to other agents.

### Can I prevent the planner agent from activating automatically?

The repository design emphasizes proactive planner activation for feature work. To bypass the planner, you would need to modify the rule set in [`rules/agents.md`](https://github.com/WorldFlowAI/everything-claude-code/blob/main/rules/agents.md) or structure requests to avoid triggering the "feature" or "refactor" classification patterns in the dispatch logic.

### How does the planner agent know what planning approach to use?

The agent reads its own metadata header in [`agents/planner.md`](https://github.com/WorldFlowAI/everything-claude-code/blob/main/agents/planner.md) to understand its scope, then processes the specific request details provided. It combines its self-described expertise with the actual user input to generate implementation plans tailored to each feature request.