rescue

Delegate complex investigation, diagnosis, or implementation tasks to Codex via the rescue subagent.

Updated Mar 26, 2026
One-click install
npx skills add https://github.com/mrkhachaturov/rkstack --skill rescue-mrkhachaturov
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: rescue
Source: https://github.com/mrkhachaturov/rkstack/tree/main/skills/rescue
Command: npx skills add https://github.com/mrkhachaturov/rkstack --skill rescue-mrkhachaturov

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Delegates substantial investigation, diagnosis, or implementation tasks to Codex via the rescue subagent when Claude is stuck, needs a second implementation pass, or should offload long-running work, reducing time-to-solution and cognitive load.

Core Features & Use Cases

  • Background execution and long-running task support through the codex-rescue flow.
  • Model selection and thread resume/fresh options to continue or start new rescue threads.
  • Clear task routing and result handling with structured outputs.

Quick Start

Invoke the rescue command with your task description to delegate work to Codex.

Frequently Asked Questions about rescue

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

FAQPage Schema
How do I delegate complex debugging tasks to an AI agent for deep investigation?

The rescue command supports background execution, allowing you to offload long-running implementation and diagnosis tasks to Codex while preserving your current working context for other operations.

Can I resume a previous Codex thread when offloading a stalled implementation task?

Yes, the rescue command supports thread resume functionality, allowing you to continue a previous Codex rescue thread with full context rather than starting a fresh investigation from scratch.

Do I need a specific subagent installed to delegate long-running work to Codex?

Yes, the codex-rescue subagent dependency is required, as it handles the underlying task routing, model selection, and result propagation between your environment and Codex.

What is the best way to offload a second implementation pass when my AI agent is stuck?

Delegating the task to Codex via the rescue subagent is the best way to handle a second implementation pass, as it routes the complex work away from your stuck agent for a deep fix.

Can I select a specific model when routing diagnosis tasks to Codex?

Yes, the rescue command includes model selection options during task delegation, allowing you to choose the appropriate Codex model for your specific diagnosis or implementation requirements.