Tech debt inventory
Leadership wants a tech-debt inventory and nobody wants to compile it by hand. Design an agent that scans the codebase and produces a ranked debt register. What signals feed the ranking, what counts as evidence for each claim, and how do you stop it from labeling every old file as debt? How would you validate the ranking against engineers' own judgment, and how do you present this so leadership funds the fixes rather than filing the report away?
rank_debt_register(module_records: list[str]) → list[str][["name=checkout;churn=6;complexity=30;incidents=2","name=legacy_report;churn=0;complexity=90;incidents=0","name=auth;churn=3;complexity=20;incidents=0"]]out["checkout","auth"][[]]out[][["incidents=1;name=payments;complexity=10;churn=0","name=search;churn=2;complexity=25;incidents=0"]]out["payments","search"]Turn the fuzzy goal into a bounded agent task. Say what the agent reads, what it may do on its own versus draft for a human, and what it must never touch. Then earn the trust: how you’d verify it before rollout, and the pitch (the problem, what you built, why it matters).
Agent build: design the agent in plain language (or narrate it). The coach grades the decomposition, the guardrails, the verification plan, and the pitch.