{"slug": "the-best-instruction-you-can-give-an-ai-coding-agent-when-uncertain-do-less", "title": "The Best Instruction You Can Give an AI Coding Agent: 'When Uncertain, Do Less'", "summary": "A developer has published UNIVERSAL-AGENTS.md, an open-source instruction file for AI coding agents whose guiding principle is \"When uncertain, do less, not more.\" The document's rules, including Section 25's \"A change being beneficial does not make it in scope,\" direct agents to avoid unrequested refactors, flag unrelated problems rather than fix them, ask one or two sharp clarifying questions instead of guessing, and follow an explicit priority order when goals conflict.", "body_md": "Ask a human teammate to fix a typo and you get a fixed typo.\n\nAsk an AI coding agent the same thing and you might get the typo fixed, the surrounding function \"cleaned up\", two variables renamed, and a note about how it also improved the error handling.\n\nEach change might be defensible on its own. Together they turn a ten-second review into a twenty-minute one, and you now own a diff you didn't ask for.\n\nThe agent isn't being malicious. It is being *helpful* with no idea where \"helpful\" stops.\n\n[UNIVERSAL-AGENTS.md](https://github.com/NTDevLops/UNIVERSAL-AGENTS.md) ends with one line:\n\n**When uncertain, do less, not more.**\n\nEverything else in the file is an elaboration of it. Here is what it looks like in practice.\n\nSection 25 has my favorite sentence in the whole file:\n\nA change being beneficial does not make it in scope.\n\nThat sentence is aimed squarely at the agent's strongest instinct. Most unwanted changes are *good ideas*: a cleaner name, a modern syntax, a dependency that's two versions behind. The rule acknowledges they might be improvements and still says no, because the request defines the work.\n\nSection 8 backs it up with a list of reasons that don't justify touching a file: it could be improved, its formatting could be modernized, its naming could be clearer, its dependencies could be newer, its tests could be expanded. A possible improvement is not automatically part of the requested work.\n\nThis is the interesting case. While working, the agent spots an unrelated problem. Sections 8 and 14 give a narrow path:\n\n**Fix it only if** it directly prevents the requested change from working, or you explicitly authorize the fix.\n\n**Otherwise:**\n\nIt's a deliberate trade-off. The agent doesn't smuggle in changes you'll discover during review, and it still flags problems that matter.\n\nDoing less also means not guessing. Section 2 says that if a request is ambiguous, incomplete, contradictory, or in conflict with the project, the agent should:\n\nAnd never guess requirements that materially affect the implementation.\n\nNote the word *minimum*. An agent that fires ten clarifying questions at you is just as annoying as one that guesses. The goal is one or two sharp questions that unblock the work.\n\nWhen goals conflict, Section 1 gives an explicit order: explicit user requirements, then correctness, then safety and security, then existing architecture and conventions, then minimal scope, then reuse, then maintainability, then documentation consistency.\n\nThat order answers", "url": "https://wpnews.pro/news/the-best-instruction-you-can-give-an-ai-coding-agent-when-uncertain-do-less", "canonical_source": "https://dev.to/codenamew/the-best-instruction-you-can-give-an-ai-coding-agent-when-uncertain-do-less-edk", "published_at": "2026-10-07 23:15:46+00:00", "updated_at": "2026-10-07 23:16:50.763472+00:00", "lang": "en", "topics": ["ai-agents", "ai-tools", "developer-tools", "generative-ai"], "entities": ["NTDevLops", "GitHub"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/the-best-instruction-you-can-give-an-ai-coding-agent-when-uncertain-do-less", "markdown": "https://wpnews.pro/news/the-best-instruction-you-can-give-an-ai-coding-agent-when-uncertain-do-less.md", "text": "https://wpnews.pro/news/the-best-instruction-you-can-give-an-ai-coding-agent-when-uncertain-do-less.txt", "jsonld": "https://wpnews.pro/news/the-best-instruction-you-can-give-an-ai-coding-agent-when-uncertain-do-less.jsonld"}}