Collect public-source updates with GitHub Agentic Workflows and review every write
At a glance
A developer tested a public-information workflow with Markdown definitions, agents for judgment calls, Actions for deterministic steps, and human review before changes landed. This workflow is listed under developer workflow, github actions, human review.
- Agents / tools
- GitHub Agentic Workflows, GitHub Actions, Markdown
SourceBuilding a tool for these workflows? See weekly rates.
What they did
1) Create a small GitHub Agentic Workflows experiment to collect and organize information from public sources. 2) Define the architecture, agents, skills, workflows, and outputs in Markdown so changes remain diff-friendly and reviewable. 3) Use agents for interpretation such as summarization, severity assessment, and deciding what to update; keep deterministic work such as scheduling, format conversion, and PDF generation in regular GitHub Actions. 4) Route every generated change through a review step before accepting the write.
How it went
The author said the review gate made generated updates easier to trust and helped clarify which steps benefited from agent reasoning. It was a personal learning experiment, not a production evaluation.
Worth knowing
The discussion reports no benchmark or production deployment; treat the process as a design experiment.
