This looks pretty cool. I am busy on a project at the moment but am thinking this could streamline deployment. I envision situations where, for instance, one of the code quality/security gates fail or a nested dependency, perhaps for a Docker image, is found. Instead of aborting the workflow and having to fix it and start over, I am thinking with this agentic workflow that we could add instructions for the agent to resolve the issue and then resume, perhaps saving some time and GitHub Action usage. Anyone done anything like that yet?
I got automated dependency management working! Dependabot identifies them and then Copilot updates them, runs lint, typecheck, vitests, playright e2e tests, updates the readme dates, bumps the version (patch) and deploys. Still needs more testing but looking good. I’m thinking a few browser based tests would be good to add. I’ll eventually throw out the system as a skill or something once tested but it wasn’t very hard to get set up. I also set it up so that every time I push changes, it runs an audit on the docker readme and github readme against the changelog and release notes for accuracy and consistency. Hoping it helps document maintenance/drift which is a pain point with AI as I bet you all know. It’s not tested yet.
Both the automated dependency system and the documentation drift catching system are included in my memory-journal-mcp project for now. The documentation drift system has now been successfully tested. There is also a GitHub Actions/Workflows health system, which I haven’t tested yet. It watches for workflow issues and fixes them. Check them out!
