Have the agent interview you before it writes code
A developer uses an agent to clear up vague requirements, handle small implementation choices, and keep them learning while the code is being built.
- What they did
- Before coding, they ask the agent to question them about unclear parts of the spec, one point at a time. They keep control of the important decisions, such as stack and architecture, but let the agent settle routine choices like login and other standard subsystems. They also use a skill that produces short explanations and quizzes, so each major step comes with a learning check.
- How it went
- The developer says this helps them avoid the late-night cycle of patching and rebreaking code while still keeping a solid grasp of what the system does. The agent also runs checks such as linters, type checks, pre-commit hooks, tests, formal-method traces, and a headless-browser design verifier. No numbers were given, and they note that other engineers disagree on whether deep understanding matters.
- Tips
- This works best when you separate important design calls from low-stakes implementation details. For anything that could change data, send messages, delete files, or make payments, have the agent stop and ask for approval first.
Try it yourself with Claude Code
I want to build [feature/project] in [codebase], and I want to keep understanding every part of it. Start by asking me one ambiguity at a time about the spec, and do not code until those are resolved. For decisions that are routine or low-risk, choose a sensible option yourself and list each choice before proceeding. As you build, after every major chunk, give me a brief explainer of what changed and a 3-question quiz to check my understanding. Run the relevant checks as you go: [linters/type checks/tests/hooks/verifiers]. If any step would change data, send something, delete something, or submit anything, pause and ask for my approval before doing it.
Fill in the [brackets]. Agents can make mistakes, so review anything before it is sent, paid or deleted.
Get 3 agent uses like this every morning
With the day's top AI news. Free, and one click to unsubscribe.
Related agent uses
Use review loops to turn a spec into working code
A developer fed a spec into an implementation engine, then repeated verification and review passes until the code matched the plan.
Review a feature and pass the findings to an engineer
One developer had an agent inspect a feature for problems, then handed the issue list to an engineer whose agent fixed most of them.
Connect a coding agent to an Excalidraw board
They hooked an installed coding agent to a live Excalidraw canvas so it could read the scene and edit the drawing directly.
Have coding agents test your onboarding flow
Run coding agents through your product’s onboarding and get a scorecard, fix list, and session traces.
Review Linux kernel patches before they land
Google’s Sashiko scans Linux kernel submissions and flags bugs that human reviewers missed.
Use an agent to inspect and explain COBOL on z/OS
A desktop agent connected to a mainframe to read COBOL, submit jobs, and summarize failures from spool output.