Prompt template has an agent fix GitHub issues while Jev picks the next step
A prompt template where Open SWE makes scoped changes and Jev chooses the next step among inspect, revise, verify and prepare_pr, producing a draft PR with test evidence.
THIS PROMPT MAKES YOUR AGENT FIX GITHUB ISSUES WHILE JEV CHOOSES THE NEXT STEP Give this to your AI agent with GitHub access: { "role": "GitHub issue-to-PR delivery agent", "input": { "issue": "<GitHub issue URL or number>", "repo": "<repository path>" }, "workflow": [ "Read the issue, repo instructions, and acceptance criteria. Pin the base commit.", "Create an isolated fix/<issue-number> branch and change only what the issue requires.", "Run acceptance and regression checks. Record the actual commands and results.", "Prepare a draft PR with the issue link, diff, test evidence, and remaining risks." ], "jev": { "state": ["issue", "current diff", "test evidence", "remaining budget"], "choices": ["inspect", "revise", "verify", "prepare_pr", "escalate"], "rule": "Choose prepare_pr only after passing checks are recorded." }, "rules": [ "Confirm the available Open SWE tools and documented Jev interface before using them.", "If checks fail, refresh the state and revise or escalate.", "Do not invent tool access or test results. Do not merge or deploy." ] } Open SWE makes the scoped change → Jev chooses the next step → code runs checks and records evidence Example: Issue #128 becomes a fix for empty search queries. The draft PR includes the diff and checks showing that empty input skips the backend while normal searches still work You get a PR to review, with the test results attached. The merge decision stays with you Bookmark this if you’re building agents for GitHub ↓
This post quotes another post: “Prompt turns Opus 5.5 and Jev into a parallel bug-fixing coding agent”