Auto-pick new bugs within team rules

A DeepSeek Harness agent watches one repo, follows team-written permissions, and opens work items only when the rules allow it.

Done with Custom agent· Advanced

What they did
The team wrapped a custom agent as a DeepSeek Harness bundle for a single GitHub repository. It watches GitHub, and optionally Feishu, then checks whether an approved person asked for help or whether a written rule allows it to start work. If so, it creates a task, runs in its own session and folder, and reports back in the original thread, such as with a pull request. A web dashboard shows the items still needing a human.
How it went
The project README does not report any measured outcome. It describes the code as research-only, not ready for production, and warns that it may still have bugs. It also says some ideas from the related article were left unbuilt.
Tips
The setup is involved, so the README has the coding agent read SETUP.md and proceed one short question at a time. If you use a similar flow, keep the approval step before any action that would send, merge, delete, or submit work.

Try it yourself with Devin

Help me set up an agent for [GitHub repository or issue tracker URL] that follows these team rules: [list your rules, such as who can request work, which labels it may act on, and which paths it must avoid]. Watch for new bugs or allowed requests, then decide whether the rules permit taking them on. If a bug is allowed, open a working branch, try to reproduce it, suggest a fix, and prepare a draft pull request tied to the issue. Before you do anything irreversible, such as sending, merging, deleting, or submitting, stop and ask me to approve the exact change. If the rules do not allow action, explain briefly why and leave it alone.

Fill in the [brackets]. Agents can make mistakes, so review anything before it is sent, paid or deleted.

Read the original on github.com ↗

Get 3 agent uses like this every morning

With the day's top AI news. Free, and one click to unsubscribe.