How to turn GitHub issues into PRs with Copilot
The author estimates about 25% of issue-to-Copilot PRs are mergeable as-is and about 50% after comments, while reporting setup compliance and manual verification limitations.
news.ycombinator.com
Suggested prompt
Suggested prompt written by Agents In Use, not the source author. Replace bracketed inputs and review tool permissions first.
Implement [GitHub issue] in [monorepo]. Follow [repository setup and coding rules], add relevant tests and open a PR explaining the change. Address findings from [automated reviewers] and report any checks or infrastructure you could not run.
Reach people exploring AI agents
See placements →Evidence at a glance
Sourced account- Source author
- adam_gyroscope
- Original account
- Read the source ↗
- Published
- Source link checked
- Reported result
- The author estimates about 25% of issue-to-Copilot PRs are mergeable as-is and about 50% after comments, while reporting setup compliance and manual verification limitations.
- Human involvement
- The author reviews PRs and uses automated code reviewers and pre-commit checks; full-system manual verification remains difficult.
A linked account supports attribution. It does not mean we independently reproduced its result.
Report an issue with this case
Open correction emailOpens your email app with the case link included. Nothing is sent automatically.
How the agent did it
- Opened a GitHub issue and assigned it to Copilot.
- Copilot opened a PR for the author to review later.
- Ran Cursor Bugbot, Copilot and CodeQL reviews before human review.
Adam_gyroscope describes a startup workflow built around a monorepo containing scheduled Python data workflows and two Next.js applications. GitHub handles source control and CI/CD, with deployments to GCP and Vercel. The team assigns every newly opened issue to Copilot so that each issue receives a code attempt, while unused PRs are frequently deleted.
When away from a keyboard, the author opens an issue through Slack, mobile or the web and assigns it to Copilot. Descriptions range from detailed instructions to a single sentence. Copilot opens a PR for later review. At the keyboard, the author instead uses Cursor in a Git worktree, iterates, asks for tests, reviews the changes and pushes them.
Before human review, Cursor Bugbot, Copilot and GitHub CodeQL review the code. The author then asks Copilot to address their findings. Pre-commit hooks enforce typing, formatting and tests across the repository.
The author estimates roughly 25% of issue-generated Copilot PRs are mergeable immediately, rising to about 50% after a few comments. They also describe unreliable setup-instruction compliance and difficulty running the complete application infrastructure for manual verification.
Who it's for: small engineering teams that can review agent-created PRs within an existing automated development process.
Prompt
Suggested prompt (written by Agents In Use, not the author's)
Source: Hacker News · adam_gyroscope
Frequently asked questions
What happens to unused Copilot PRs?
The team frequently deletes PRs it does not use and describes having many attempted PRs as acceptable.
What verification limitation does the author describe?
The author needs Temporal, two Next.js apps, multiple workers and a browser for manual checks. GitHub Actions only approximates that setup and does not solve manual verification or database work.
More details
Remaining steps
- Asked Copilot to fix flagged findings and iterated through comments.
About this account
Copilot produces code attempts and opens PRs from assigned issues, then responds to requested fixes and review comments.
Platform: Hacker News. Attribution: https://news.ycombinator.com/item?id=46292682. Prompt: Suggested (written by Agents In Use, not the author's). Sourced by the Agents In Use pipeline.
Credit: adam_gyroscope on Hacker News
Source published
Added to Agents In Use
Source checked
Tools and topics
GitHub Copilot · GitHub · Cursor Bugbot · CodeQL · Pre-commit hooks
GitHub issues · Code review · Monorepo
Editorial completeness: 100/100. Measures documented fields, not agent performance or verified outcomes.