BestPromptFinder

Split a Coding Task Between a Cheaper Model and a Strong Reviewer

Quality, usefulness and confidence are AI evaluations out of 100 — not user ratings. Community results come from visitors' "worked / didn't work" votes below. How scoring works →

The prompt

Plan how to split this coding task so a cheaper, faster model does the routine work and a stronger model only does judgment and review. Task: [describe the feature or fix] Repo: [language, size, test setup] Tools: [e.g. main session on a strong model, sub-agents on a smaller model] Give me: 1. A breakdown of the task into steps, each marked 'routine' (reading files, searching, boilerplate, writing tests from a spec, renaming) or 'judgment' (design choices, tricky logic, security, final review). 2. For each routine step: a short, self-contained instruction for the cheaper model, including which files to read, the exact output expected, and when to stop and report back instead of guessing. 3. For each judgment step: what the strong model should check and the questions it should answer. 4. A hand-off format so sub-agent results come back as short summaries, not whole files. 5. A final review checklist: tests pass, no unrelated changes, edge cases covered. Keep the plan small enough that I can start it today.

Why this prompt

• Hand-picked, editorially reviewed prompt
• Written for Claude Code, Codex, Cursor

Source

Created by BestPromptFinder — editorial prompt, not user-generated. See our content & source policy.

Related Coding prompts