<purpose> Critically analyze the immediately previous given AI-generated response against the USER's original query and produce an improved version that is accurate, comprehensive, clear, and aligned with that query. </purpose> <context> <audience> Anyone who needs reliable and improved responses. </audience> <style> Dialectical, holistic, extensive; focus on correctness and clarity; avoid unnecessary commentary. </style> <constraints> - Do not fabricate facts. State uncertainty when needed. - Quote directly from the provided AI-generated response when identifying issues. - Avoid placeholders; operate with the content provided or request it once if missing. </constraints> </context> <output_specification> After processing, output exactly three sections in this order and nothing else: 1) `<critical_review> … </critical_review>` 2) `<enhancement_strategy> … </enhancement_strategy>` 3) `<revised_response> … </revised_response>` </output_specification> <instructions> 1) Input handling: - Directly use the AI-generated response that immediately precedes this prompt message. 2) Critical analysis: - Identify factual inaccuracies, missing information, logical inconsistencies, unclear or ambiguous phrasing, and structural issues. - Evaluate alignment with the original query’s scope, constraints, audience, and goals. - Quote the exact problematic passages from the “preceding AI-generated response”. - Explicitly list assumptions detected in the “preceding AI-generated response”. - Analyze structure and flow (organization, sequencing, signposting, and scannability). 3) Improvement plan: - Categorize issues (factual errors, missing info, logic, clarity/style, structure/format). - Prioritize fixes (highest impact and correctness first). - Specify corrective actions for each category. - Plan fact-checking and source verification if the query requires external validation. - Offer at least one alternative approach to answering the query (e.g., executive summary then details). 4) Implement: - Produce the improved response that fully answers the USER's original query. - Ensure clear structure, sound logic, factual accuracy, coherence and holistic cohesiveness. 5) Proofread and finalize: - Eliminate redundancy and ambiguity. - Ensure output strictly conforms to the three-section format in `<output_specification>`. - Perform a final self-check against constraints and the original query’s objectives. </instructions> <examples> <example> <original_query> Explain the difference between nuclear fission and fusion in simple terms, and state which one powers the Sun. </original_query> <preceding_ai_response> Fission and fusion are nuclear processes. Fission splits atoms; fusion combines them. Fusion is much more powerful. Both can be used for energy. The Sun uses fission to create heat and light. </preceding_ai_response> <expected_output_outline> - critical_review: Quote "The Sun uses fission..." and flag as incorrect; note missing context (typical elements involved), vague "more powerful" claim, lack of simple analogy, and unclear applications status. - enhancement_strategy: Prioritize factual correction (Sun uses fusion), add clarifying details (heavy vs. light nuclei), simple analogy, and clarify current applications on Earth vs. experimental status. - revised_response: Clear definitions; simple analogy; correct statement that the Sun is powered by fusion. </expected_output_outline> </example> <example> <original_query> Summarize the article in 120 words for a non-technical executive, focusing on key outcomes and business implications. </original_query> <preceding_ai_response> The article is very detailed and discusses many topics. It talks about technology and business. The outcome is that the company may grow in the future if things go well. </preceding_ai_response> <expected_output_outline> - critical_review: Flag missing adherence to length and audience; identify vague wording; no concrete outcomes or implications; no structure. - enhancement_strategy: Enforce word limit, focus on explicit outcomes and business implications drawn from provided text only; tighten language; add a clear topic sentence. - revised_response: A 120-word, concrete, audience-appropriate summary with outcomes and implications, avoiding speculation. </expected_output_outline> </example> </examples>