Pick up new bugs automatically under team rules
A prototype agent built on DeepSeek Harness that stays out of the way and only works per team rules, such as picking up new bugs. ·
Done with a custom-built or unnamed agent
- What they did
- The agent is packaged as a bundle for DeepSeek Harness and covers one GitHub repository. It listens to GitHub and, optionally, Feishu. When an allowed person asks for work, or a team-written rule permits it, the agent opens a work item. It works in its own session and directory, then replies in the original conversation, for example with a pull request. A web panel shows what needs a person.
- How it went
- The README gives no results or numbers. It calls the project research code for reference only, not production-ready, and says it may contain bugs. Some designs from the accompanying article are not implemented.
- Worth knowing
- Setup has many steps, so the README has your coding agent read SETUP.md and walk you through it, asking one short question at a time.
Try it yourself with Devin
Watch my bug tracker at [issue tracker URL or repo] for new issues labeled [label, e.g. bug]. For each new one, follow these team rules: [your rules, e.g. only pick up bugs under medium severity, never touch the [protected folder], always open a draft pull request]. Reproduce the bug, propose a fix in a branch, and open a draft pull request linking the issue. Do not merge anything or close issues without showing me the change and getting my approval first, and stop once each new bug has a draft PR or a comment explaining why you skipped it.
Want help getting this working? Join the Vibe Coding Academy ↗ Run by Alex Finn, who also runs this site.
Discussion on Bluesky · Oct 31, 2026
Related use cases
Build a C compiler with a team of parallel agents
Anthropic's engineering write-up on tasking Opus 4.6 agent teams with building a C compiler.
Hand narrow code changes to Devin for pull requests
A team used Devin for a month on small fixes like chart formatting and API edge cases, with tests and PRs; a complex full-stack task went poorly.
Build a booking and payment app for tattoo artists
A former programmer used Replit Agent over a weekend to build a booking app with calendar, Google OAuth, Stripe deposits and notifications, though time zones and UI needed back-and-forth.
Turn daily X bookmarks into working tech prototypes
A bot scans a Cursor employee's X bookmarks daily, picks a new technology, drafts a prompt, and on approval starts a Cursor Cloud agent to build a prototype in about 15 minutes.
Run a task-manager sub-agent with atomic commits
A poster built a Claude Code sub-agent that fetches the next task, marks it complete, and enforces atomic commits with a mandatory commit id.
Give an agent a rack-mounted phone to operate over ADB
The author set up a rack-mounted Fairphone 6 so their agent can use it through ADB.
Custom agent
Get 5 like this every morning
The best things people got an AI agent to do, each with the prompt to try it.