read-only

Provides codebase analysis, debugging, and in-chat code solutions without any filesystem mutations.

1|Updated Apr 20, 2026
One-click install
npx skills add https://github.com/ntd237/draft_toolkits_20042026 --skill read-only-ntd237
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: read-only
Source: https://github.com/ntd237/draft_toolkits_20042026/tree/main/.agents/skills/read-only
Command: npx skills add https://github.com/ntd237/draft_toolkits_20042026 --skill read-only-ntd237

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? When you need technical guidance, debugging help, or codebase analysis but cannot risk an AI agent accidentally creating, editing, or deleting files, this Skill enforces an absolute zero-disk-mutation policy so all solutions arrive as copy-paste-ready code in chat. ## Core Features & Use Cases - Zero-Disk-Mutation Enforcement: Blocks all write, edit, and delete tools across AI platforms (Claude Code, Cursor, Codex, Cline, Roo Code, Windsurf, Antigravity) plus mutating shell commands like rm, sed -i, and git reset --hard. - Safe Inspection & Diagnosis: Uses read-only tools (grep, file viewers, git status, git diff, Playwright, web search) to gather evidence and perform root cause analysis. - In-Chat Code Delivery: Outputs complete, syntactically valid code blocks or unified diffs with exact file paths and line numbers for manual application. - Use Case: You ask the AI to fix a crash in src/utils/pricing.ts. Instead of editing the file, it reads the code, identifies a division-by-zero bug, and returns the corrected function in chat with step-by-step application instructions. ## Quick Start Ask the AI to analyze a bug or review a file in your project while explicitly requesting that no files be modified on disk.

Frequently Asked Questions about read-only

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I get AI code fixes without the agent modifying my files?▼

Activate a read-only policy that blocks all write, edit, and delete tools so the agent only inspects code and returns complete fixes as chat code blocks. You then copy the suggested code into the specified file and line range manually.

Which AI coding agents support a read-only mode?▼

This policy covers Claude Code, Cursor, Codex, Cline, Roo Code, Windsurf, and Antigravity by mapping each platform's mutation tools (Edit, Write, apply_patch, replace_in_file) to a prohibited list. Any custom MCP write tools are also blocked.

Can the AI still run terminal commands in read-only mode?▼

Yes, but only inspect-only commands such as git status, git log, git diff, ls, cat, grep, curl, and dry-run checks. Mutating commands like rm, sed -i, git reset --hard, and file redirection are prohibited.

What happens if I ask the read-only agent to edit a file?▼

The agent reminds you of the active read-only policy and instead delivers the complete, ready-to-use code directly in chat. It includes the exact file path, line numbers, and step-by-step instructions for applying the change yourself.

What are the limitations of a read-only AI workflow?▼

The agent cannot apply changes, run builds that write artifacts, or create temporary scratch files, so every fix requires manual copy-paste by the user. This adds a manual step but guarantees the filesystem stays untouched.