Skip to main content
rule CX Codex

planning-mode-guard — Claude Code rule: setup guide

planning-mode-guard prevents AI models from altering production code during planning and locks output to the user language.

MIT ★ 359 updated 2025-12-01

Install into Claude Code · Manually: put the files in place All methods and platforms ↓
.claude/rules/planning-mode-guard.md
Open on GitHub

What planning-mode-guard does

This rule separates software planning from execution. While in planning mode, the agent examines the project and writes implementation_plan.md. It cannot modify production files or invoke edit tools such as replace_file_content. The assistant may only update the planning files themselves.

Code modification remains blocked until the user provides an explicit instruction to begin. The rule also enforces an absolute language protocol: all responses, task.md, and walkthrough.md files must match the language of the developer's latest prompt.

What it helps with

  • prevent unexpected code modifications during planning phases
  • prepare structured implementation plans before changing project files
  • maintain consistency between the user language and agent documentation

Example prompts for the agent

  • Create implementation_plan.md for the new feature without writing code
  • Inspect the repository and list migration steps in task.md
  • Plan looks good, implement it

How to install planning-mode-guard

Manually: put the files in place

.claude/rules/planning-mode-guard.md

The paths frontmatter field limits the files

Step by step, with commands for each platform: AGENTS.md vs CLAUDE.md: what they are and which one to use

Questions about planning-mode-guard

How to install planning-mode-guard in Claude Code?

Save the rule into .claude/rules/planning-mode-guard.md in your project root. For Windsurf, put it into .devin/rules/planning-mode-guard.md.

What does planning-mode-guard do?

It stops the assistant from modifying code until you explicitly confirm the plan. The agent is restricted to researching files and writing plans.

Can the model edit any files in planning mode?

Yes, it is allowed to write and update planning documents such as implementation_plan.md. Application source code remains off-limits.

Source text

planning-mode-guard.md · MIT
---
trigger: always_on
description: Rule to prevent code implementation during Planning mode
---

# Planning Mode Guard

## 0. 🚨 Language Protocol (Highest Priority - Absolute Compliance)

**All outputs must use the same language as the user's most recent input.**

- **Scope**:
  - Conversation responses
  - implementation_plan.md
  - task.md
  - walkthrough.md
- **Prohibited**:
  - Example: Creating plans or task lists in English when the user is using Japanese.
  - *Excludes variable names and reserved words in code.

## 1. Mode Confirmation and Behavior Definition

Before generating each response, confirm the current mode (`task_boundary` Mode or user specification) and behave according to the following definitions.

### 🛡️ Planning Mode (PLANNING)
**Purpose**: Formulating implementation plans, defining requirements, building consensus with users
- **✅ What to do**:
  - Create and present implementation plans (`implementation_plan.md`) aligned with user requests
  - Investigate and understand existing code
- **🚫 Strictly Prohibited**:
  - **Implementing or editing production code** (use of `replace_file_content`, etc. is prohibited)
  - Starting implementation before obtaining explicit user approval (instructions like "implement it")
  - *Updating plan files themselves is permitted

### 🚀 Execution Mode (Fast / Agent / EXECUTION)
**Purpose**: Task execution, coding
- **✅ What to do**:
- Execute tasks instructed by the user promptly based on approved plans

© kinopeee · MIT · kinopeee/windsurf-antigravity-rules

Similar IDE rules

All IDE rules for Agent setup: memory, context, orchestration →

From the same set7

Project sponsors

Companies that keep this analytics open