phx-freeze

Applies an advisory session scope limiting which project files the agent may edit.

537|38|Updated Feb 12, 2026
One-click install
npx skills add https://github.com/oliver-kriska/claude-elixir-phoenix --skill phx-freeze
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: phx-freeze
Source: https://github.com/oliver-kriska/claude-elixir-phoenix/tree/main/targets/codex/skills/phx-freeze
Command: npx skills add https://github.com/oliver-kriska/claude-elixir-phoenix --skill phx-freeze

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

When working on a focused task in an Elixir/Phoenix codebase, you may want the AI agent to avoid touching files outside a specific area. This Skill establishes a session-level instruction that restricts edits to read-only mode or to specific directory prefixes, reducing the risk of unintended changes.

Core Features & Use Cases

  • Read-Only Mode: Invoke with no arguments to prevent any file edits while keeping investigation and reporting available.
  • Directory-Scoped Editing: Pass project-relative path prefixes (e.g., lib/app_web priv/repo) to allow edits only under those directories.
  • Scope Management: Use status to report the current advisory scope and off to clear it.
  • Use Case: While investigating a bug in the web layer, freeze edits to lib/app_web so the agent can fix the controller without accidentally modifying contexts, schemas, or migrations.

Quick Start

Ask the agent to freeze edits to the lib/app_web directory using the phx-freeze skill before starting the investigation.

Frequently Asked Questions about phx-freeze

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

FAQPage Schema
How do I restrict an AI agent to edit only certain directories?

Invoke phx-freeze with project-relative path prefixes, such as `lib/app_web priv/repo`. The agent will then edit only files under those directories and ask before touching anything outside them.

How do I make an AI coding session read-only?

Invoke phx-freeze with no arguments. The agent will not edit any files, while investigation, analysis, and reporting remain fully available for the session.

Is the phx-freeze edit scope enforced by the runtime?

No, the scope is advisory only. This generated runtime installs no enforcement hook, so it is a binding instruction for the current agent rather than a technical lock or security boundary.

How do I check or clear the current freeze scope?

Run phx-freeze with `status` to report the advisory scope active in the conversation, or with `off` to clear it for subsequent work in the session.

Does phx-freeze create a .claude/.freeze file?

No, it must not create or modify `.claude/.freeze`. That sentinel belongs to the canonical Claude Code plugin, and this runtime cannot enforce or clear it reliably.