Skip to content
Musher Docs

Export Integrations

The Musher SDK can export bundle content into formats understood by Claude, OpenAI, VS Code, and other platforms. This lets you pull a bundle once and feed its skills, prompts, and toolsets directly into your AI toolchain.

Claude Plugin Export

Export a bundle's skills as a Claude Desktop plugin. This creates the directory structure Claude expects, with a manifest and skill files:

python
import musher

bundle = musher.pull("acme/code-review-kit")

# Export all skills as a Claude plugin
result = bundle.export_claude_plugin("code-review")
print(f"Plugin created at: {result.path}")
print(f"Plugin name: {result.plugin_name}")

# Export specific skills only
result = bundle.export_claude_plugin(
    "security-only",
    skills=["security-scan"],
)

Claude Skill Install

Install skills directly into a Claude configuration directory. This is useful for scripting local development setups:

python
import musher
from pathlib import Path

bundle = musher.pull("acme/code-review-kit")

# Install all skills to a destination
bundle.install_claude_skills(dest=Path("~/.claude/skills").expanduser())

# Install specific skills, removing existing ones first
bundle.install_claude_skills(
    dest=Path("~/.claude/skills").expanduser(),
    skills=["code-review"],
    clean=True,
)

OpenAI Local Skill

Export a skill as an OpenAI-compatible local skill definition. The result is written to disk, ready to register with OpenAI's API:

python
import musher

bundle = musher.pull("acme/code-review-kit")
skill = bundle.skill("code-review")

local_skill = skill.export_openai_local_skill()
print(local_skill.name)
print(local_skill.description)
print(local_skill.path)

# Convert to a dict for API submission
payload = local_skill.to_dict()

OpenAI Inline Skill

Export a skill as an inline skill with base64-encoded content. This embeds the skill content directly in the API request, avoiding the need for file paths:

python
import musher

bundle = musher.pull("acme/code-review-kit")
skill = bundle.skill("code-review")

inline_skill = skill.export_openai_inline_skill()
print(inline_skill.name)
print(inline_skill.description)
print(inline_skill.content_base64)

# Convert to a dict for API submission
payload = inline_skill.to_dict()

VS Code Skill Install

Install skills into a VS Code agents directory. This writes skill files to .agents/skills/ where VS Code extensions can discover them:

typescript
import { pull, installVSCodeSkills } from "@musher-dev/musher-sdk";

const bundle = await pull("acme/code-review-kit");

// Install skills to .agents/skills/
await installVSCodeSkills(bundle);

VS Code integration is currently available in the TypeScript SDK only.

Export Types Reference

Python

TypePropertiesSource method
ClaudePluginExport.plugin_name, .pathbundle.export_claude_plugin()
OpenAILocalSkill.name, .description, .pathskill.export_openai_local_skill()
OpenAIInlineSkill.name, .description, .content_base64skill.export_openai_inline_skill()

TypeScript

FunctionDescription
exportClaudePlugin(bundle)Creates .claude-plugin/plugin.json directory structure
installClaudeSkills(bundle)Writes skills to .claude/skills/
exportOpenAILocalSkill(bundle)Writes skill as local file
exportOpenAIInlineSkill(bundle)Creates base64-encoded ZIP for inline use
installVSCodeSkills(bundle)Writes skills to .agents/skills/