[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"glossary-meta-prompt::en":3,"gloss-cluster-meta-prompt::en":20,"gloss-next-meta-prompt::en":9},{"slug":4,"category":5,"name":6,"definition":7,"meta_desc":8,"faq":9,"schema_markup":9,"related":10},"meta-prompt","prompt-eng","Meta Prompt","A meta prompt is a prompt used not to directly accomplish an end-user-facing task, but to generate, improve, evaluate, or otherwise operate on other prompts — using an LLM's own language capabilities to assist in the prompt-engineering process itself. This \"prompting about prompting\" pattern has become increasingly practical as models have gotten better at understanding what makes an effective instruction, and several frontier labs now publish or expose meta-prompting tools (Anthropic's prompt generator and prompt improver in the Claude Console, OpenAI's prompt-generation tooling) that take a rough task description and produce a well-structured, best-practices-following starter prompt. Meta prompts show up in a few distinct production patterns: prompt generation (given a task description, generate a first-draft production prompt following known best practices — clear role, explicit format instructions, few-shot placeholders); prompt critique\u002Fimprovement (given an existing prompt and examples of where it underperforms, suggest specific edits); and LLM-as-judge evaluation (a meta prompt that scores or compares the outputs of other prompts against a rubric, used to automate prompt A\u002FB testing at a scale manual review can't match — e.g., \"Given this task description and these two candidate responses, which better satisfies the criteria: accuracy, conciseness, tone? Respond with A, B, or TIE and a one-sentence justification\"). For SaaS builders running many AI features, meta-prompting is a genuine productivity multiplier in the prompt-engineering workflow, especially LLM-as-judge patterns for automated evaluation — instead of a human manually reviewing hundreds of test-set outputs after every prompt tweak, a well-calibrated judge meta-prompt can score them automatically, letting teams iterate on prompts far faster while still spot-checking judge accuracy against human review periodically. Concrete worked example: a team maintaining 40 different AI features across their SaaS product uses a meta prompt to evaluate every prompt change before deploy: \"You are evaluating two versions of an AI feature's output for the same input. TASK: {{task_description}}. INPUT: {{test_input}}. VERSION A OUTPUT: {{output_a}}. VERSION B OUTPUT: {{output_b}}. Score each 1-10 on accuracy and format-compliance, then state which version is better overall and why.\" Run automatically across a 200-example regression test set on every prompt-template pull request, this meta prompt catches regressions (a prompt edit that improved one scenario but broke another) before they reach production, functioning as an automated CI check for prompt quality. Meta prompting has limits worth flagging: a model generating or judging prompts is still subject to the same failure modes (biases, occasional misjudgment, sensitivity to how the meta-task itself is phrased) as any other LLM call, so teams typically treat an AI-generated or AI-improved prompt as a strong first draft that a human still reviews, and calibrate an LLM-as-judge's scoring against a sample of real human ratings periodically rather than trusting it as an infallible ground truth from day one.","A meta prompt is a prompt whose job is to generate, refine, or evaluate other prompts, rather than answer an end-user task directly.",null,[11,14,17],{"slug":12,"name":13},"prompt-engineering","Prompt Engineering",{"slug":15,"name":16},"prompt-library","Prompt Library",{"slug":18,"name":19},"prompt-versioning","Prompt Versioning",[21,25,28,31,35,38,41,44,47,50,53,56],{"slug":22,"category":5,"name":23,"updated_at":24},"analogical-prompting","Analogical Prompting","2026-08-24T02:46:37+00:00",{"slug":26,"category":5,"name":27,"updated_at":24},"automatic-prompt-optimization","Automatic Prompt Optimization",{"slug":29,"category":5,"name":30,"updated_at":24},"chain-of-density","Chain of Density (CoD)",{"slug":32,"category":5,"name":33,"updated_at":34},"chain-of-thought-prompting","Chain-of-Thought Prompting","2026-08-24T02:46:36+00:00",{"slug":36,"category":5,"name":37,"updated_at":24},"chain-of-verification","Chain-of-Verification",{"slug":39,"category":5,"name":40,"updated_at":34},"chunking","Chunking",{"slug":42,"category":5,"name":43,"updated_at":34},"constrained-decoding","Constrained Decoding",{"slug":45,"category":5,"name":46,"updated_at":34},"context-stuffing","Context Stuffing",{"slug":48,"category":5,"name":49,"updated_at":34},"delimiter","Delimiter",{"slug":51,"category":5,"name":52,"updated_at":24},"directional-stimulus-prompting","Directional Stimulus Prompting",{"slug":54,"category":5,"name":55,"updated_at":24},"emotion-prompting","Emotion Prompting",{"slug":57,"category":5,"name":58,"updated_at":34},"few-shot-prompting","Few-Shot Prompting"]