Generated from docs/why.md at df1dbe1. Edit the repository file, not the wiki.

Why

One maintainer with evenings and weekends, a dozen repositories, and a team of AI agents: a Claude lead with Claude subagents, OpenAI Codex, and GitHub Copilot's coding agent. Agents multiply output. Without a lead, review gates and one place to see the work, they multiply unverified churn instead.

The board is that one place

A GitHub Project is where the maintainer looks to answer three questions: what is being worked on, by whom (which agent), and what is waiting on a person. That only works if the board is complete (every open issue and pull request of every repository is on it), current (Done means closed or merged, with the date), and cheap to keep that way (nobody drags cards).

What safo is for

• Board as code. board.yaml names the fields, options, iterations and views. safo bootstrap creates what is missing and refuses to edit what exists, because the API call that edits options wipes every item's value.
• An Action that keeps it current. sync reacts to one issue or pull request event; reconcile repairs what an event missed, on a schedule; audit tells you when the live board and the file disagree.
• A playbook. Roles, routing, handoff and limits for a lead, its subagents and the other agents, each with a rule the lessons earned.

What it will not do

It does not run agents. It does not host anything: the Action runs in your workflow with your App's key. It does not touch fields it was not told about, and it never reopens a card.

The shape of a good week

The maintainer decides and merges. The lead breaks work into cards, sets the Agent field, hands each task out with a brief, and checks the result with a fresh reviewer. CI green is the bar for a merge; the board shows the rest.