{"slug": "use-blunt-prompts-and-get-shit-done", "title": "Use Blunt Prompts and Get Shit Done", "summary": "A developer has shared a blunt prompting technique for coding agents that demands a 20% minimum increase in test coverage while requiring the agent to find at least one real regression. The approach blocks \"coverage theater\" by forcing agents to hunt for proof that code can break, rather than simply executing lines. The developer argues that useful tests catch broken behavior or expose edge cases, while weak tests that only assert a result is not null produce noise, not confidence.", "body_md": "Do not ask a coding agent for “more tests.”\n\nThat is how you get coverage theater.\n\nAsk for pressure:\n\n```\nIncrease test coverage by at least 20%.\n\nIf you can’t find at least one real regression, your tests are shit anyway.\n```\n\nThat prompt works because it has two parts.\n\nFirst, it gives the agent a measurable floor:\n\n```\nIncrease test coverage by at least 20%.\n```\n\nNot “add some tests.”\n\nNot “improve coverage.”\n\nAt least 20%.\n\nThe number matters because it removes the cheap escape hatch.\n\nSecond, it blocks fake success:\n\n```\nIf you can’t find at least one real regression...\n```\n\nThat changes the job.\n\nThe agent is no longer writing tests to satisfy a report.\n\nIt is hunting for proof that the code can break.\n\nThat is the point.\n\nCoverage alone only says:\n\nThis line was executed.\n\nRegression hunting asks:\n\nDid this test prove anything useful?\n\nA weak test can execute code and still protect nothing:\n\n```\nself::assertNotNull($result);\n```\n\nThat is not confidence.\n\nThat is noise.\n\nA useful test catches broken behavior, exposes a missing edge case, kills a mutant, or forces you to clarify what the code is supposed to do.\n\nThe real question is:\n\nWhat did the new tests actually discover?\n\nIf the answer is “nothing,” keep going.", "url": "https://wpnews.pro/news/use-blunt-prompts-and-get-shit-done", "canonical_source": "https://dev.to/suckup_de/use-blunt-prompts-and-get-shit-done-4mdd", "published_at": "2026-05-26 22:40:20+00:00", "updated_at": "2026-05-26 23:03:41.037284+00:00", "lang": "en", "topics": ["ai-agents", "artificial-intelligence", "ai-tools", "ai-products"], "entities": [], "alternates": {"html": "https://wpnews.pro/news/use-blunt-prompts-and-get-shit-done", "markdown": "https://wpnews.pro/news/use-blunt-prompts-and-get-shit-done.md", "text": "https://wpnews.pro/news/use-blunt-prompts-and-get-shit-done.txt", "jsonld": "https://wpnews.pro/news/use-blunt-prompts-and-get-shit-done.jsonld"}}