Code
Find your tech debt
Where it is, measured, rather than where people say it is.
Fill it in
Where it is.
Complex code nobody touches is not the problem.
What to pay down first and why.
Your prompt
Find the debt in [the repository], with evidence. Compute per file: complexity, size, change frequency, number of distinct authors, and where the bugs landed. Weight by CHURN. Complex code nobody has touched in three years is not debt, it is furniture; complex code that changes every week is where the cost is. Ranking on complexity alone sends teams to refactor a stable parser while the file that causes every incident stays as it is. Cross-reference the hotspots against the bug history. A file that is complex, changes constantly and appears in half the incidents is the report. Propose an order, with a size estimate and what improves for each. "Refactor the checkout module" is not a plan. Say which hotspots are inherent complexity rather than accidental.
Use Find your tech debtOpens with everything above already filled in.
Why this works
Ranking on complexity alone sends teams to refactor a stable parser while the file that causes every incident stays as it is. Complex code nobody touches is furniture. This weights by churn and cross-references the bug history.