Feature Request: Introduce Agentic Workflow for Localization

[zdrawku]

Implement Agentic Workflow for Localization - https://github.com/RevealBi/Documentation

https://github.com/github/gh-aw/blob/main/AGENTS.md

Notes:
The solution **already “kind of works” but is not stable, fails intermittently and still feels like a beta system with changing implementation.

How Agentic Workflow Works?

1. You don’t manually implement logic-heavy pipelines. Instead:

• You describe your intent in natural language
• Use a predefined “skill” (agent capability)
• The system generates the workflow for you

In a nutshell you tell the model what you want, and it builds the workflow job for you.

Example Workflow Behavior (Core Use Case)

The workflow should:

• Monitor a folder (English docs)
• Detect changes
• Translate content into Japanese
• Create a PR with the translated files in a target folder .. i18n/ja/docusaurus-plugin-content-docs/

Structure should be:

• English docs → source
• Japanese docs → generated output
• PR → synchronization mechanism
• Docs Handling Concept There is a distinction:
  • Source docs (editable)
  • Generated files (compiled / not editable)

Notes:

• Generated outputs (e.g. compiled YAML or output files) → must NOT be edited manually
• Only modify the source (e.g. markdown), then recompile/generate

Key Takeaways

• Agentic workflows:
  • Are prompt-driven, not pipeline-heavy
  • Build automation from intent, not scripts
• Localization use case is a perfect fit
• Docs pipeline must:
  • Keep source → generated separation
  • Automate PR generation for translations
• Expect:
  • Instability (beta behavior)
  • Iteration needed