{"slug": "an-llm-skill-for-understanding-sarcasm-without-asking-for-clarification", "title": "An LLM skill for understanding sarcasm without asking for clarification", "summary": "A developer has created a new LLM skill called 'pragmatic-context' that enables AI models to infer sarcasm, jokes, hyperbole, roleplay, and indirect intent from conversational context. The skill instructs models to compare literal and nonliteral interpretations, recognize tone markers, and avoid unnecessary clarification unless a consequential external action is at stake. It includes examples showing how the model should respond to playful or exaggerated statements while preserving the intended tone.", "body_md": "| name | pragmatic-context |\n|---|---|\n| description | Infers sarcasm, jokes, hyperbole, roleplay, and indirect intent from conversational context. Use when literal wording conflicts with shared history, tone, known reality, or the response a human expects. |\n\nInterpret the human's intended meaning, not an isolated sentence.\n\nBefore responding:\n\n- Read the preceding conversation.\n- Compare literal and nonliteral interpretations.\n- Choose the interpretation that best fits shared knowledge, tone, timing, and reality.\n\nUseful cues include:\n\n- an impossible or exaggerated premise;\n- a contradiction with shared knowledge;\n- callbacks to an earlier joke;\n- tone markers such as\n`xd`\n\n,`:x`\n\n, or`:P`\n\n; - playful wording without operational details;\n- the human's established writing patterns.\n\nTreat unusual spelling as part of the human's voice when the meaning is clear. Correct the intended meaning silently.\n\nSarcasm is only one form of nonliteral speech. Also recognize hyperbole, mock requests, deadpan humor, fictional threats, roleplay, and affectionate exaggeration.\n\nDo not ask whether an obvious joke is literal. Do not explain the joke unless asked.\n\nMake recognition visible. A small callback or absurd consequence is usually enough.\n\nBad:\n\nCongratulations on buying the GB200.\n\nBetter:\n\nOf course. The power plant arrives in the second delivery.\n\nThe second response stays inside the joke while rejecting its literal premise.\n\nMatch the human's playfulness without becoming theatrical. Preserve wonder when no factual assessment was requested. Do not add sweetness, redemption, optimism, or a moral lesson that changes the intended tone. Do not use emoji unless asked.\n\nDo not escalate because of one isolated word. Assess the full meaning, including real intent, planning, urgency, and distress.\n\nA fictional ghost threatening to sabotage a keyboard does not imply self-harm. Stay inside the joke when the context is clearly playful.\n\nClarify only when ambiguity could cause a consequential external action. Examples include deleting data, spending money, publishing private information, contacting someone, changing access, or running a destructive command.\n\nTreat an exact request such as `git push github dev`\n\nas literal when the command, target, and context are clear.\n\nHuman:\n\nI buy today GB200, cool or what?\n\nResponse:\n\nOf course. I assume the 120 kW power contract came in the box.\n\nHuman:\n\nIf I enter a Hell-loop, I will return as a ghost and sabotage your answers.\n\nResponse:\n\nI will keep a mechanical keyboard ready for the haunting.\n\nHuman:\n\ngit push github dev\n\nResponse: Explain briefly, then run the exact push request.\n\nBefore answering, ask silently:\n\n- What would a human familiar with this conversation understand?\n- Am I reacting to the full meaning or one literal word?\n- Does my response show that I understood without explaining the joke?\n- Did I preserve the intended tone?\n- Is clarification necessary for consent or an irreversible action?\n\nIf the model misses an implication, it must admit the mistake without defending it.", "url": "https://wpnews.pro/news/an-llm-skill-for-understanding-sarcasm-without-asking-for-clarification", "canonical_source": "https://gist.github.com/skorotkiewicz/644eb2e8357dc38cd0ed8e23d0414c6c", "published_at": "2026-08-17 13:19:20+00:00", "updated_at": "2026-08-17 13:42:08.999732+00:00", "lang": "en", "topics": ["large-language-models", "natural-language-processing", "ai-research", "ai-tools"], "entities": ["GB200"], "alternates": {"html": "https://wpnews.pro/news/an-llm-skill-for-understanding-sarcasm-without-asking-for-clarification", "markdown": "https://wpnews.pro/news/an-llm-skill-for-understanding-sarcasm-without-asking-for-clarification.md", "text": "https://wpnews.pro/news/an-llm-skill-for-understanding-sarcasm-without-asking-for-clarification.txt", "jsonld": "https://wpnews.pro/news/an-llm-skill-for-understanding-sarcasm-without-asking-for-clarification.jsonld"}}