BestPromptFinder

Safety Brief for an AI Coding Agent: Budget, Loops and Stop Rules

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

Add these working rules to the top of my coding agent's instructions (CLAUDE.md / AGENTS.md / project rules). Rewrite them to fit my project, keep them short and specific. Project: [what it is, language, how tests run] Paid services the code can call: [e.g. an LLM API, cloud functions, SMS] My spending limit for a single session: [amount] Rules to include: 1. Never start a background process, loop or scheduled job that calls a paid API without telling me the expected number of calls and the cost, and waiting for my OK. 2. Every loop or retry has a hard cap (max iterations and max minutes), written in the code. 3. Test against mocks or the cheapest model by default; real paid calls only when I ask. 4. Before ending a turn, list any process still running and either stop it or tell me exactly how to stop it. 5. Never report a task as idle or done while something it started is still running. 6. If a command runs longer than [10] minutes, stop and report. Then give me a 5-line checklist I can paste into each new session.

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