DeepSeek open-sources a team coding agent that takes work from GitHub and only logs what it would do until you allow it

DeepSeek published dsh-superagent-prototype on GitHub on Oct 31 under the MIT license. It is a plugin bundle for DeepSeek Harness that listens to one GitHub repository and, optionally, Feishu. A main router session decides whether a message is a request; an accepted request becomes a Ticket that runs in its own worker session and directory and answers in the original thread, for example with a pull request. By default, only people on an allowed list can start work by addressing it, and proactive work starts in an `observe` mode that only records what it would have done. DeepSeek calls it a research prototype, not an official product, and says it differs from its internal version. Why it matters: An agent that works for a whole team can also waste the whole team's attention, or act on a request from the wrong person. DeepSeek's prototype puts the rules in writing. Only people on an allowed list can give it work, and comments in a thread can't change its permissions. Any work it would start on its own just gets logged until someone reviews the log and switches that on. DeepSeek calls it a research prototype, not a product.

Try this: Open docs/002-authorization.md in the repo and write down, for your own repo, who could ask an agent for work and which work it may do unasked. Use that list as the allowed list before you install anything.