Prompt-craft & iterating

🎯Goal. Get sharper results by scoping the task, giving Codex something to verify against, and steering as it works.

1. 1Be concrete about the goal and the guardrails: "Find and fix bugs in my codebase with minimal, high-confidence changes" beats "fix everything".
2. 2Give it a target to check itself against — a failing test, expected output, or a screenshot (attach images with -i / --image or by pasting them).
3. 3Let it explore first ("Tell me about this project") before asking for changes, then refine across turns — Codex keeps the thread's context.

✓You'll see. The difference between a vague ask that over-reaches and a scoped one that makes the minimal, correct change.

💳Cost. Scoped prompts and a clear "done" signal mean fewer model round-trips — less of your plan's usage per task.

💡Takeaway. Tell Codex exactly what success looks like and give it a way to verify — it iterates toward a target far better than toward a vibe.