The Bug Fix template makes your AI coding agent resolve a bug the same way every time: check the report is actionable, reproduce it, find the root cause, apply a surgical fix with a regression test, then verify. It stops twice for you, and a failed verification sends it back to Fix.
6 steps · 2 approvals · 1 loop back
example · the steps are this template's real steps
For engineering teams who fix bugs with Claude Code, Codex, Cursor or OpenCode and want the reproduction and the root cause on record, not in a chat log.
It is a stored workflow, not a prompt. Your AI client receives one step at a time from ConvOps over MCP, does the work, and advances. The engine holds the run at each of its two approvals until a person says yes.
6 steps, in this order. Below each one is the instruction your AI receives at that step, exactly as the template stores it.
Verify the bug report is actionable before investigation begins.
Required elements:
If any element is missing, identify the gap and ask the operator to clarify. Do not advance with an under-specified report.
Produce a minimal, deterministic reproduction of the reported behavior.
Capture:
If the bug cannot be reproduced, stop and report that finding explicitly. Do not proceed on speculation: an irreproducible report requires clarification, not guesswork.
Request operator confirmation that this reproduction matches the reported issue before advancing.
no rejection path · the run moves on only when approved
Identify the root cause and scope the smallest corrective change.
Document:
Keep the analysis focused and concise: depth of reasoning matters more than length.
Apply the corrective change defined in the analyze step. Keep the edit surgical: do not refactor adjacent code, rename unrelated symbols, or expand scope beyond the root cause.
Add or update a regression test that would have caught this bug before the fix. The test must fail on the unfixed code and pass on the fixed code.
Verify the fix resolves the reported behavior without introducing regressions.
Execute:
Report what was executed, what passed, and what failed. If any check fails, reject this step: the workflow returns to fix so the gap can be addressed before another verification pass.
Request operator approval before advancing.
rejected → back to Fix
The bug has been reproduced, analyzed, fixed, and verified. Mark the task completed and close it out.
Two steps wait for a person. You answer in the chat; the engine records the verdict and the audit row records who gave it.
A fix built on the wrong reproduction fixes the wrong thing. You confirm the recipe matches the reported issue before any analysis starts.
what you review
no rejection path · approve to continue
The reviewer reports what ran, what passed and what failed. You decide whether the fix closes. Reject it and the run goes back to Fix.
what you review
reject → back to Fix
gates the engine checks without you
Read from the template's own steps: the tools they call, the files they write, and whether they touch code.
Connect your AI client first, then install Bug Fix from the catalog in the ConvOps app. See how to connect a client
An installed template is a workflow in your workspace. Ask your AI to change it in plain words; the edit lands on the one step that needs it.
new chat
more example prompts
The Reproduce step tells your AI to stop and report that finding explicitly instead of guessing. An irreproducible report goes back to you for clarification, and the run waits at the Reproduce approval until you have confirmed a reproduction.
Yes. The Fix step asks for a regression test that fails on the unfixed code and passes on the fixed code, and Verify runs it together with the reproduction recipe and the existing test suite at the relevant level.
The reviewer rejects the Verify step and the workflow returns to Fix. The gap is fixed there and verification runs again, so a failed check never closes the bug.
Bug. Once installed, a new bug in your workspace starts on this workflow, and your AI receives Intake as its first step.