clone_skill

Overview

This skill clones the current Claude conversation so you can branch off and try a different approach without losing the original thread. It uses your local Claude history to locate the session and a helper script to create a timestamped cloned conversation you can reopen with the CLI. The result is a separate session labeled with a [CLONED <timestamp>] tag for easy discovery.

How this skill works

The skill reads the latest history entry to extract the session ID and project path, then finds the helper script clone-conversation.sh in your Claude configuration directory. It runs the script with the session ID and the recorded project path (not your current working directory). The script creates a cloned conversation and timestamps it; you reopen clones using claude -r and look for the [CLONED <timestamp>] marker.

When to use it

• You want to experiment with a different line of questioning without losing the original conversation.
• You need a safe copy before trying risky edits or automated transformations.
• You want to compare multiple approaches or prompts side-by-side.
• You are debugging prompt variations and need reproducible branches.
• You want to preserve a stable baseline before applying new tools or plugins.

Best practices

• Always use the project path taken from the history record, not your current directory, to ensure the clone references the correct context.
• Prefer the latest clone-conversation.sh by locating it with version-sorted find to avoid running outdated scripts.
• Confirm the cloned session appears in claude -r and check the [CLONED <timestamp>] label before proceeding.
• Keep a convention for naming or tagging clones in your notes so you can track experiments across sessions.
• Run clones before making bulk edits or running agentic workflows that change state.

Example use cases

• Branching off an architecture discussion to test a different design pattern while keeping the original thread intact.
• Creating a copy of a debugging session before running experimental patches or automated refactors.
• Testing multiple prompt strategies for a coding task and comparing outputs without losing any history.
• Safely handing a cloned session to a teammate for parallel work without affecting your active conversation.
• Preserving a conversational baseline before integrating new plugins or automations.