The agent can receive GitHub events and connect to Feishu. After an authorized request, it works in a separate session and directory, then returns the result to the conversation where the request began; it can respond if blocked by a decision requiring human input.
Packaged for DeepSeek Harness, the project accompanies the company’s paper on super-agent design. DeepSeek says coding agents authored part of the code and warns it may contain bugs and inconsistent engineering practices.
Why it matters
The prototype offers a concrete example of a repository-scoped coding workflow designed to reduce human attention, but it is not ready for production.
Something to try
As a hypothetical evaluation, try a simple task in a test repository and note where the agent requests human input.
Start the conversation.