Agentic Workspaces

What counts as a workspace.

An agentic workspace is a repository you work inside with your agent. The agent is given tools, rules, skills and scripts that steer it and help it produce better, more consistent output. What you produce lands directly in the repo: your applications, notes, research, drafts.

What an agentic workspace is An agent session reads tools, guidance and constraints from a job-specific folder, writes artifacts back into it, and reads those artifacts as context on the next run. agent session competitor-research/ › read AGENTS.md · tools/ · constraints.md › read artifacts/ 14 notes · 6 research › draft artifacts/drafts/v3.md › log artifacts/decisions.md reads writes competitor-research/ AGENTS.md tools/ constraints.md artifacts/ research/ notes/ drafts/ decisions.md … 6 14 3 SET UP ONCE what the agent may do GROWS EVERY RUN what the work left behind next run · the agent reads these too

The four rules

  1. It ships a workspace you use as a project directory — either an assembled folder structure or an installation script or skill.
  2. You operate it primarily through your own agent.
  3. State accumulates. Your data lands in the workspace and becomes part of its context.
  4. The output is your work — what the workspace is designed to produce: applications, papers, campaigns, decisions.

Propose a workspace

Send one this list is missing. It goes through the four rules above, judged by the same classifier a crawl result meets.

One sentence is plenty.

How the list was built

A crawler enumerates candidates by repository search and by citation, and every one is judged against the four rules above. The crawl isn't perfect, and we can't be certain every result is what we were looking for.