Extract specific projects from a Claude.ai data export ZIP file. Lists all projects in the export, lets the user pick which to extract, and saves project knowledge docs and related conversations to chosen directories. Trigger with /ClaudeProjectExport. Use when the user has a Claude.ai export ZIP and wants to pull project data to their local drive, mentions "claude export", "extract projects from export", or wants to migrate Claude.ai project context into Claude Code.

Claude Export Extractor — Claude Code Skill

Extract projects from a Claude.ai data export ZIP into local directories for use with Claude Code.

What This Skill Does

Claude.ai projects contain knowledge docs, prompt templates, and conversation history. This skill extracts specific projects from an export ZIP into organized local folders so the content can be used as context for Claude Code work.

Output structure per project:

<output_dir>/
  project_knowledge/          # Knowledge docs, attachments, prompt template
    _project_metadata.json    # Project name, UUID, dates, counts
    _prompt_template.md       # Project custom instructions (if any)
    <knowledge files>...      # All uploaded docs, deduplicated
  conversations/              # Related conversation history as markdown
    <conversation>.md ...     # One file per conversation

Workflow

Step 1: Locate the ZIP

Ask the user for the path to their Claude.ai export ZIP file. Common locations:

~/Downloads/
C:\Downloads\
The user may have already mentioned it

Verify the file exists before proceeding.

Step 2: List projects

Run the extractor in JSON mode to get a machine-readable project list:

PYTHONIOENCODING=utf-8 python ~/.claude/skills/ClaudeProjectExport/claude_export_extractor.py "<zip_path>" --json

This returns a JSON array with each project's name, doc count, conversation count, and size.

Step 3: Present projects to the user

Show the projects in a clean table format. Ask which projects to extract.

Step 4: Get output directories

For each selected project, ask where to save the output. Suggest a sensible default based on the project name and current working directory.

Step 5: Extract

Run the extractor in non-interactive mode:

PYTHONIOENCODING=utf-8 python ~/.claude/skills/ClaudeProjectExport/claude_export_extractor.py "<zip_path>" --extract <nums> --output "<dir1>,<dir2>"

Step 6: Report results

Confirm what was saved: doc count, conversation count, output paths.

Notes

Conversations are matched to projects by keyword similarity (no UUID link in export format)
Duplicate docs within a project are automatically deduplicated by filename
PYTHONIOENCODING=utf-8 is required on Windows to avoid emoji encoding errors
Pass all numbers to --extract for a full dump