Claude Skill Porter

Export Claude Code plugin knowledge to Gemini, ChatGPT, Claude Projects

A CLI tool that extracts expert skills from Anthropic's knowledge-work-plugins (52 skills across 11 plugins) and exports them to formats compatible with Gemini CLI, ChatGPT Custom Instructions, and Claude Projects.

---

Quick Start

Get up and running in 30 seconds:

# 1. Export a skill to Gemini CLI format
export-skill product-management feature-spec gemini > feature-spec-gemini.txt

# 2. Export a skill to ChatGPT Custom Instructions
export-skill data sql-queries chatgpt

# 3. Export all skills from a plugin to Claude Projects format
export-plugin marketing ~/exports/marketing claude-project

Expected output:

• Single skill exports: Formatted content written to stdout or file
• Plugin exports: Directory created with one file per skill
• Success messages: ✓ Exported [skill-name] to [format]

---

Installation

Option 1: Curl One-Liner (Recommended)

curl -fsSL https://raw.githubusercontent.com/tentaclequing/claude-skill-porter/main/install.sh | bash

This installs export-skill and export-plugin to ~/.local/bin (added to PATH automatically).

Option 2: Manual Installation

# Clone repository
git clone https://github.com/tentaclequing/claude-skill-porter.git
cd claude-skill-porter

# Add bin/ directory to PATH
echo 'export PATH="$HOME/claude-skill-porter/bin:$PATH"' >> ~/.bashrc
source ~/.bashrc

Verify Installation

export-skill --version
# Expected: claude-skill-porter v1.0.0

---

Usage

List Available Plugins

export-skill --list-plugins

Output:

Available plugins:
  - content-creator
  - data
  - marketing
  - product-management
  - sales
  - [... 6 more]

List Skills in a Plugin

export-skill product-management --list-skills

Output:

Available skills in 'product-management':
  - feature-spec
  - metrics-tracking
  - roadmap-planning
  - user-story

Export a Single Skill

Syntax:

export-skill <plugin-name> <skill-name> <format>

Formats: gemini, chatgpt, claude-project

Examples:

# Export to Gemini CLI (writes to stdout)
export-skill data sql-queries gemini > sql-skill-gemini.txt

# Export to ChatGPT Custom Instructions (creates two files)
export-skill product-management feature-spec chatgpt
# Creates: feature-spec-chatgpt-field1.txt (role context)
#          feature-spec-chatgpt-field2.txt (instructions)
# Note: ChatGPT has 1500-char limit per field

# Export to Claude Projects (markdown format)
export-skill marketing campaign-planning claude-project

Export Entire Plugin (Batch Export)

Syntax:

export-plugin <plugin-name> <output-dir> <format>

Example:

export-plugin marketing ~/exports/marketing gemini

Output:

✓ Exported campaign-planning to gemini (~/exports/marketing/campaign-planning-gemini.txt)
✓ Exported social-media to gemini (~/exports/marketing/social-media-gemini.txt)
✓ Exported content-calendar to gemini (~/exports/marketing/content-calendar-gemini.txt)

Batch export complete: 3 skills exported to ~/exports/marketing

---

Format Comparison

Format	Character Limit	Best Use Case	Example Command	Notes
Gemini CLI	None	System instructions for Gemini CLI	export-skill data sql-queries gemini	Full content preserved, wrapped with role context
ChatGPT Custom Instructions	1500 chars/field	ChatGPT web UI settings	export-skill marketing campaign-planning chatgpt	Smart truncation for large skills, overflow file created
Claude Projects	None	Claude Projects knowledge base	export-skill product-management feature-spec claude-project	Clean markdown, includes attribution

Format-Specific Details

Gemini CLI

• Output: Single .txt file with system instruction wrapper
• Structure: Role context header + full skill content
• Usage: Copy content into Gemini CLI system instruction parameter
• Character limit: None

ChatGPT Custom Instructions

• Output: Two files per skill (*-field1.txt, *-field2.txt)
• Structure:
  • Field 1 (150-300 chars): Role context from plugin/skill description
  • Field 2 (1400-1500 chars): Core content with smart truncation
• Truncation strategy: Preserves key principles, structure, first 1-2 examples
• Overflow handling: If content > 3000 chars, creates *-extended.md with full content
• Usage: Paste field1 into "What would you like ChatGPT to know about you?" and field2 into "How would you like ChatGPT to respond?"

Claude Projects

• Output: Single .md file
• Structure: Clean markdown with frontmatter removed, attribution footer
• Usage: Upload to Claude Projects knowledge base
• Character limit: None

---

Advanced Usage

Using Custom Plugin Directory

By default, claude-skill-porter looks for plugins at:

~/.claude/plugins/marketplaces/knowledge-work-plugins/

If your plugins are located elsewhere, use the --plugin-dir flag:

# List plugins from custom directory
export-skill --list-plugins --plugin-dir /path/to/custom/plugins

# Export skill from custom directory
export-skill data sql-queries gemini --plugin-dir /path/to/custom/plugins

LLM-Assisted Command Generation

Not sure which plugins or skills are available? Or need help constructing the right command? Use this prompt template with Gemini, ChatGPT, or Claude:

I'm using claude-skill-porter to export Claude Code plugin skills to your format.

My plugin directory is at: [PASTE YOUR PATH HERE - default is ~/.claude/plugins/marketplaces/knowledge-work-plugins/]

Please help me:
1. Verify this directory exists and contains plugin folders
2. List all available plugins and their skills
3. Generate the correct export command for the skill I want

I want to export: [DESCRIBE WHAT YOU'RE LOOKING FOR]
Examples:
- "a skill about writing SQL queries"
- "all marketing-related skills"
- "the feature spec template from product management"

How this works:

1. Paste the prompt into your AI assistant
2. Replace [PASTE YOUR PATH HERE] with your actual plugin directory path
3. Describe what skill you're looking for
4. The AI will guide you to the right command

Example interaction:

You: [Paste prompt above, filling in path and "a skill about writing SQL queries"]

AI: I can see your plugins directory. The 'data' plugin contains a 'sql' skill.
     Here's your command:

     export-skill data sql-queries gemini > sql-skill-gemini.txt

     This will export the SQL query skill to Gemini CLI format.

Version Tagging

All exported content includes source version information in the header/footer:

<!-- Exported from knowledge-work-plugins v1.2.0 on 2026-02-09 -->

This helps you track which version of the source material was used, so you can re-export if skills are updated.

---

Troubleshooting

Error: Plugin directory not found

Symptom:

ERROR: Plugin directory not found: ~/.claude/plugins/marketplaces/knowledge-work-plugins/

Solution:

1. Verify Claude Code is installed and plugins are synced:

   ls ~/.claude/plugins/marketplaces/knowledge-work-plugins/

2. If plugins are elsewhere, use --plugin-dir flag:

   export-skill --list-plugins --plugin-dir /your/custom/path

Error: Plugin 'xyz' not found

Symptom:

ERROR: Plugin 'xyz' not found

Available plugins:
  - content-creator
  - data
  [...]

Solution:

1. Check available plugins: export-skill --list-plugins
2. Verify spelling (plugin names are case-sensitive and hyphenated)
3. Common mistakes:
   • productmanagement → product-management
   • DataAnalyst → data

Error: Skill 'xyz' not found in plugin 'abc'

Symptom:

ERROR: Skill 'xyz' not found in plugin 'abc'

Available skills in 'abc':
  - skill-one
  - skill-two
  [...]

Solution:

1. List skills in plugin: export-skill abc --list-skills
2. Verify skill name spelling
3. Common mistakes:
   • featureSpec → feature-spec
   • SQL → sql

Error: Cannot read SKILL.md file

Symptom:

ERROR: Cannot read SKILL.md file: /path/to/plugin/skills/xyz/SKILL.md

Solution:

1. Check file permissions:

   ls -la ~/.claude/plugins/marketplaces/knowledge-work-plugins/plugin-name/skills/skill-name/

2. Ensure file exists and is readable:

   cat ~/.claude/plugins/marketplaces/knowledge-work-plugins/plugin-name/skills/skill-name/SKILL.md

3. If file is missing, re-sync Claude Code plugins

ChatGPT Export Warning: Content Truncated

Symptom:

WARNING: Content truncated to fit ChatGPT 1500-char limit
See feature-spec-extended.md for full content

Solution: This is expected behavior for large skills. The truncated version preserves key content, while the -extended.md file contains the complete original content. You can:

1. Use the truncated version in ChatGPT Custom Instructions
2. Reference the extended file in your ChatGPT conversations when needed
3. Or use Gemini/Claude Projects formats which have no character limits

Permission Denied Error

Symptom:

bash: export-skill: Permission denied

Solution:

1. Make scripts executable:

   chmod +x ~/claude-skill-porter/bin/export-skill
   chmod +x ~/claude-skill-porter/bin/export-plugin

2. Or reinstall using curl method (sets permissions automatically)

---

Contributing

Contributions are welcome! Here's how to get started:

Adding a New Format

1. Create a new format converter in lib/formats/<format-name>.sh
2. Implement the format_<name>() function with this signature:

   format_my_format() {
     local skill_file="$1"    # Path to SKILL.md
     local plugin_name="$2"   # Plugin name
     local skill_name="$3"    # Skill name
   
     # Your conversion logic here
     # Output formatted content to stdout
   }

3. Source your converter in bin/export-skill and bin/export-plugin
4. Add tests in tests/test-<format-name>.sh
5. Update this README with format documentation

Example: See lib/formats/gemini.sh for a simple reference implementation.

Testing Guidelines

• All format converters must have test coverage
• Tests should verify:
  • Correct output format
  • Character limits (if applicable)
  • Error handling for missing/invalid inputs
  • Edge cases (empty content, special characters)
• Run tests before submitting PR:

  ./tests/run-all-tests.sh

Code Style Guidelines

Bash scripting conventions:

• Use set -euo pipefail at top of all scripts
• Quote all variable expansions: "$variable" not $variable
• Use local for function variables
• Prefer [[ over [ for conditionals
• Use $(command) over backticks
• Functions should output to stdout, errors to stderr
• Use return 1 for errors, return 0 for success
• Document functions with comments explaining purpose and parameters

Naming conventions:

• Files: kebab-case.sh
• Functions: snake_case()
• Variables: snake_case for locals, SCREAMING_SNAKE_CASE for globals

Testing:

• Test files: test-<feature>.sh
• Use descriptive test names: test_format_gemini_basic_export()
• Include setup/teardown for temp files

Submitting a Pull Request

1. Fork the repository
2. Create a feature branch: git checkout -b feature/my-new-format
3. Make your changes following style guidelines
4. Add/update tests
5. Run test suite: ./tests/run-all-tests.sh
6. Update README if adding user-facing features
7. Commit with clear message: Add support for Ollama format
8. Push and create PR with description of changes

---

License

Licensed under the Apache License, Version 2.0. See LICENSE for full text.

Source Material Attribution

This tool extracts knowledge from Anthropic's knowledge-work-plugins, which is also licensed under Apache 2.0. All exported content retains its original licensing.

See THIRD_PARTY_NOTICES for detailed attribution.

---

Roadmap

v1.0 (Current):

• ✓ Export to Gemini CLI, ChatGPT, Claude Projects
• ✓ Single skill and batch plugin export
• ✓ Custom plugin directory support
• ✓ Smart truncation for ChatGPT

v1.5 (Planned):

• Local LLM formats (Ollama, LM Studio)
• Reverse import (other AI tools → Claude skills)
• Auto-sync with knowledge-work-plugins updates
• Interactive mode with fuzzy search

v2.0 (Future):

• Web catalog UI for browsing and exporting skills
• CI/CD pipeline for auto-exporting on plugin updates
• Custom skill creation tools integration

---

Support

• Issues: GitHub Issues
• Discussions: GitHub Discussions
• Source: knowledge-work-plugins

---

Built with ❤️ for the multi-tool AI practitioner community