AgenticImport

AgenticImport uses AI agents to migrate workflows between repositories with automatic adaptation. Unlike gh aw add or shared imports, it creates independent, customized copies without version tracking.

When to Use

Use AgenticImport for one-time migrations requiring adaptation:

Repository forking - Adapt workflows for different projects
Context-specific changes - Substantial modifications for new environments
Cross-organization - Copy workflows where shared imports aren’t practical
Experimentation - Fork workflows as automation starting points

Don’t use for synchronized updates—use Packaging & Distribution with gh aw add instead.

How It Works

The create-agentic-agent migrates workflows through these steps:

Analyze source - Read and understand the source workflow’s purpose and logic
Identify dependencies - Detect repository-specific configuration, secrets, tools, and permissions
Adapt configuration - Modify frontmatter, imports, and instructions for target repository
Create workflow - Generate or update the workflow file in the target repository
Validate - Compile and suggest fixes for any issues

Example Migration

The release.md workflow migration from githubnext/gh-aw to githubnext/gh-aw-mcpg:

Migrate the release.md workflow from githubnext/gh-aw to this repository.

The workflow should:
- Keep the same release automation logic
- Adapt permissions and jobs for our repository structure
- Update any repository-specific references
- Ensure all imports and tools are compatible

The agent automatically handles repository names, paths, permissions, and tool configurations.

Setup

Initialize workflow creation agents:

gh aw init

AgenticImport vs Traditional Import

Use AgenticImport for:

Complex adaptations (repository-specific config, tool changes)
One-time migration without ongoing synchronization
AI-assisted analysis and validation

Use gh aw add for:

Synchronized updates across repositories
Version management with semantic tags
Minimal customization

Use manual copy for:

Simple workflows without complex dependencies

Advanced Patterns

Merging Workflows

Merge issue-triage from org/repo1 and label-management from org/repo2
into a unified workflow handling both responsibilities.

Splitting Workflows

Split multi-responsibility workflow.md into:
- Issue triage (issue-triage.md)
- PR review (pr-review.md)
- Security scanning (security-scan.md)

Extract shared configuration into common imports.

Cross-Organization

Migrate ci-doctor from public-org/workflows to our private enterprise
repository, adapting for internal MCP servers and security policies.

Real-World Example

The release.md migration from githubnext/gh-aw to githubnext/gh-aw-mcpg preserved release automation logic while adapting permissions, repository references, and structure for the target environment.

Best Practices

Before Migration:

Review source workflow purpose and dependencies
Identify required secrets, tools, and permissions
Understand shared imports
Plan necessary adaptations

During Migration:

Provide target repository context
Specify security policies and constraints
Use clear, specific instructions
Validate incrementally with gh aw compile

After Migration:

Test in safe environment (TrialOps)
Review and document changes
Remove source: frontmatter field

Authoring with AI

In GitHub Copilot Chat, you can use custom agents to help create and debug workflows:

Type /agent and select create-agentic-workflow for interactive workflow creation/migration
Type /agent and select debug-agentic-workflow for compilation debugging

Limitations

Point-in-time copies - Not for continuous synchronization (use gh aw add)
Requires context - Agent needs clear target repository instructions
Manual validation - Review and test before production
Iteration needed - Complex workflows may need refinement

Packaging & Distribution - Shared workflows with version management
Custom Agents - Specialized workflow authoring
TrialOps - Safe workflow testing
Workflow Structure - Format reference