Which migration agent does this relate to?
Is this a bug or an improvement?
Describe the issue or improvement
The current Reusable Workflow Builder agent requires GitHub organization names as input and performs broad scanning across all repositories in those organizations to discover CI/CD patterns. This approach has limitations:
- Overly broad scope — Scanning entire organizations pulls in repositories that may be irrelevant to the desired reusable workflows, adding noise to the pattern analysis.
- Lack of intent — The agent infers what reusable workflows to build solely from pattern frequency, with no way for the user to express what they actually need.
- Slow discovery — Enumerating and analyzing all repositories in large organizations is time-consuming when the user already knows which repositories are representative.
The updated agent should change its input model to accept:
- A list of repositories (e.g.,
org/repo1, org/repo2) instead of organization names. This gives users precise control over which codebases are analyzed.
- High-level requirements describing the type of reusable workflow(s) to build (e.g., "Create a reusable workflow for Node.js CI with linting, testing, and Docker image publishing" or "Build a standardized Terraform deployment workflow"). This lets users steer the output rather than relying entirely on automated pattern detection.
The agent should use the provided repositories as reference implementations — analyzing their existing CI/CD configurations to extract concrete patterns, versions, and configurations — then combine that analysis with the user's stated requirements to generate targeted reusable workflows.
Additional context
- The current agent prompt expects input in the form:
"Analyze organizations: [org1, org2, org3] for CI/CD patterns". The new input format should be something like: "Build reusable workflows from repositories: [org/repo1, org/repo2] — Requirements: [description of desired workflows]".
- The agent should still support analyzing CI/CD files from multiple platforms (GitHub Actions, GitLab CI, Jenkins, Azure DevOps, etc.) within the specified repositories.
- Pattern frequency analysis should be scoped to the provided repository list only, not expanded to the broader organization.
- All existing output rules (reusable workflow files in
.github/workflows/reusable-*.yml, usage docs in docs/<name>-usage.md, no scripts, no README files) should remain unchanged.
- The high-level requirements should guide which patterns the agent prioritizes and what the generated workflows focus on, while the repository analysis provides the concrete implementation details.
Which migration agent does this relate to?
Is this a bug or an improvement?
Describe the issue or improvement
The current Reusable Workflow Builder agent requires GitHub organization names as input and performs broad scanning across all repositories in those organizations to discover CI/CD patterns. This approach has limitations:
The updated agent should change its input model to accept:
org/repo1,org/repo2) instead of organization names. This gives users precise control over which codebases are analyzed.The agent should use the provided repositories as reference implementations — analyzing their existing CI/CD configurations to extract concrete patterns, versions, and configurations — then combine that analysis with the user's stated requirements to generate targeted reusable workflows.
Additional context
"Analyze organizations: [org1, org2, org3] for CI/CD patterns". The new input format should be something like:"Build reusable workflows from repositories: [org/repo1, org/repo2] — Requirements: [description of desired workflows]"..github/workflows/reusable-*.yml, usage docs indocs/<name>-usage.md, no scripts, no README files) should remain unchanged.