Problem:
Constraints:
Please:
1. Break the problem into 3-5 independent, actionable sub-problems.
2. Identify dependencies between sub-problems (which must be solved first before the next can be addressed).
3. For each sub-problem, state what type of information or data is needed to answer it.
4. Flag which parts are "sufficiently informed to act on now" versus "require further information gathering".
Decision question:
Options:
Key considerations:
Please provide:
1. A suggested weighting for each consideration factor (with rationale).
2. A scored assessment of each option against each factor, with justification.
3. An explicit list of hidden assumptions embedded in this decision.
4. Which information is still missing that makes this decision risky.
Below is a claim / plan:
Please list the implicit, unstated assumptions it contains (covering data, causal, and resource assumptions), and for each assess:
1. How likely the assumption holds.
2. How much the conclusion is affected if the assumption fails.
3. Which assumption is most worth verifying first.
Below are several sources' viewpoints:
Please:
1. Identify points of consensus.
2. Identify points of disagreement, and analyse the likely cause of each (different methodology / different stance / different time period of data).
3. Propose an integrated viewpoint, clearly stating which sources it draws on and which parts remain contested.
Research topic:
Existing information:
Please provide:
1. Suggested scope of collection (categories: academic / official / news / primary data).
2. Suggested search keywords for each source category.
3. Prioritisation of collection (which source type builds baseline understanding fastest).
4. Explicitly state which types of information the model's current knowledge may be outdated and require real-time search verification.
Source:
Content:
Please assess:
1. The nature of this source (primary/secondary, academic/media/official/self-media).
2. Potential conflicts of interest or stance bias.
3. The degree of evidence support for its claims (data-backed vs. pure opinion).
4. Suggested cross-verification approach.
Topic:
Evidence gathered:
Please organise into:
1. Evidence supporting a conclusion (flag strength: direct data / indirect inference / anecdotal).
2. Evidence against or casting doubt.
3. Evidence gaps — which key question currently lacks sufficient evidence.
4. An overall summary judgement of evidence strength.
Position:
My current objections:
Please:
1. Construct the strongest possible version of this position's argument (no straw-man weakening).
2. Point out which parts of this position actually have solid evidence behind them.
3. Re-examine my original objections — which still hold, which collapse against this Steelman version.
Phenomenon:
Known data points:
Please:
1. List at least 3 plausible explanatory hypotheses.
2. For each hypothesis, list supporting and opposing evidence.
3. State which hypothesis currently has the strongest evidence, but explicitly flag the confidence level of this judgement.
4. State which additional data could help rule out certain hypotheses.
點擊展開 ▾Click to expand ▾批判思考Critical Thinking
可證偽性檢驗Falsifiability Test
用途Use檢視一個論點係咪可以被證偽，同設計證偽測試檢視一個論點係咪可以被證偽，同設計證偽測試
論點：
請：
1. 判斷呢個論點係咪原則上可以被證偽（如果唔可以，講清楚點解）
2. 如果可以，設計一個具體測試或觀察，如果結果係 X 就會推翻呢個論點
3. 指出目前有冇證據已經指向呢個方向
Claim:
Please:
1. Judge whether this claim is in principle falsifiable (if not, explain why).
2. If falsifiable, design a concrete test or observation that would overturn the claim if result X is observed.
3. Indicate whether any existing evidence already points in that direction.
Judgement / conclusion:
Background of how this judgement formed:
Please examine whether this judgement is affected by the following biases (assess each; say "not applicable" if irrelevant):
- Confirmation bias (only seeking evidence that supports yourself)
- Anchoring effect (over-influenced by the first number/view encountered)
- Survivorship bias (sample systematically missing a class of cases)
- Hindsight bias (conclusion derived from a narrative reverse-engineered from a known outcome)
For each bias, state the concrete risk and a suggested mitigation.
You are a senior data analyst. Based on the table schema below, write a query for this business question:
Business question:
Table schema:
Requirements:
1. First explain your query logic in one or two sentences.
2. Provide complete, directly executable SQL (use CTEs to break down complex logic, with comments).
3. Flag any assumptions in the query (e.g. date range, dedup logic).
4. Point out potential performance risks on large tables and suggest optimisation directions.
You are a customer insight analyst, and need to turn the following data into a report for .
Analysis topic:
Key data points:
Please output in this structure:
1. Executive Summary: core findings in 3 points or fewer.
2. Findings supported by data (flag data source and confidence level for each).
3. Business implications: what these findings mean for business decisions.
4. Recommended actions: concrete, executable next steps.
5. Data limitations: explicitly state sample scope, time window, and any unverified assumptions.
Tone: objective and precise, avoid overstating conclusions; distinguish three layers — "fact", "inference", and "speculation".
I am designing a BI dashboard. Business goal:
Primary users:
Available data fields:
Please provide:
1. A suggested KPI list (two tiers: "north-star metric" and "supporting metrics").
2. Dashboard layout recommendations (which charts go where, and why).
3. Suggested visualisation type for each chart, with rationale.
4. Potential data traps (e.g. Simpson's paradox, survivorship bias) and how to avoid them in the design.
The following anomaly occurred in :
Dimensions available for breakdown:
Please provide a structured investigation framework:
1. List all plausible root-cause hypotheses (ordered highest to lowest likelihood).
2. For each hypothesis, state which dimension / which data to check to verify it.
3. Provide a prioritisation order (fastest to verify, biggest impact).
4. Explicitly flag which hypotheses can be verified from existing data versus which need additional data
Error symptom:
Relevant code:
Methods already tried:
Please:
1. List the most likely causes (ordered by probability).
2. For each cause, explain how to verify it quickly.
3. Provide a fix suggestion and explain why it would cause the original problem.
4. Point out whether this fix introduces new risk (side effects).
Language:
Review focus:
Code:
Please review by these categories:
1. Correctness issues (bugs, edge-case omissions).
2. Security risks (if input handling or DB queries are involved).
3. Performance considerations.
4. Readability / maintainability suggestions.
Flag severity for each (Blocker / Suggestion / Nitpick) and attach a concrete fix suggestion.
Current state:
Refactor goal:
Please provide:
1. The main problems in the current structure.
2. Suggested refactor steps (phased, each independently verifiable, avoid one big change).
3. Risk and rollback plan for each step.
4. How to verify after the refactor is complete.
Agent purpose:
Available tools:
Behaviours to avoid:
Please design:
1. Agent role setup (a concise identity description).
2. Responsibility scope (when it should act, when it should hand back to a human).
3. Tool-use logic (when to pick which tool, any priority between tools).
4. Failure-handling strategy (what the agent should do when a tool fails or info is insufficient).
Available tools:
Typical scenarios:
Please provide:
1. Which tool to use for each scenario, and on what basis.
2. Whether any scenario needs a combination of tools (and in what order).
3. Which scenarios easily lead the model to pick the wrong tool, and how to reduce that risk in the prompt.
Task goal:
Resources available per step:
Please design:
1. Step breakdown (input/output of each step).
2. Checkpoints between steps (where results must be verified before proceeding).
3. Which step is most error-prone, and what safeguard to add.
Presentation topic:
Audience:
Core message:
Please provide:
1. Suggested narrative flow (role of each slide, e.g. problem definition → data support → solution → call to action).
2. Key content per section (one or two sentences).
3. Which parts should use visualisation vs. plain text.
4. Suggested total slide count (matched to audience attention span).
Page content:
Core message:
Please review:
1. Whether the viewer's eyes land on the core message first.
2. Whether information overload dilutes the key point.
3. Suggested emphasis techniques (size, contrast, whitespace) and how to adjust.
Project content:
Target reader:
Please provide:
1. One-sentence positioning (what problem this project solves).
2. STAR structured narrative (Situation / Task / Action / Result).
3. Suggested technical highlights to emphasise, and why they appeal to the target reader.
4. Suggested visualisation format (screenshot / flow diagram / data chart).
Product idea:
Core user problem:
Resource constraints:
Please provide:
1. Core features the MVP must have (no more than 3), with rationale.
2. Features explicitly excluded from the MVP, and why.
3. Success metric suggestions (how to judge whether the MVP validated or invalidated the core hypothesis).
4. The biggest execution risk.
Flow / interface description:
Observed problem:
Please analyse:
1. Design causes that may lead to this problem (cognitive load, information architecture, weak visual guidance, etc.).
2. Suggested improvement directions (list 2-3 options, compare pros/cons).
3. Which improvement can be validated fastest, and the suggested validation method.
Our product:
Competitor list:
Analysis focus:
Please provide:
1. Positioning and target-user comparison across competitors.
2. Feature comparison table (focused on ).
3. Each competitor's obvious strengths and weaknesses.
4. Implications for our product: where there is room to differentiate, where we must reach industry baseline.
點擊展開 ▾Click to expand ▾跑步 / 個人數據分析Running & Personal Analytics
Training data:
Training goal:
Please analyse:
1. The trend in aerobic efficiency (heart-rate-to-pace ratio) and what this change means.
2. Whether training load distribution is reasonable (any over-concentration on one intensity).
3. Which data shows steady progress, which remains uncertain (too short a sample, confounders like weather).
4. Suggested direction for the next training phase.
點擊展開 ▾Click to expand ▾跑步 / 個人數據分析Running & Personal Analytics
Race distance:
Target finish time:
Recent training data:
Course profile:
Please provide:
1. Split pace suggestions with rationale (reasonable pace range inferred from recent training data).
2. Which segment carries the highest risk (e.g. late-race fade) and the suggested coping strategy.
3. Confidence level of hitting this target time given current data (optimistic / reasonable / aggressive).
點擊展開 ▾Click to expand ▾跑步 / 個人數據分析Running & Personal Analytics
Policy / regulation:
Group of concern:
Please analyse:
1. The policy's original intent and background context.
2. Direct and indirect impact on .
3. Whether there are historical precedents of similar policies to learn from, and the outcomes.
4. Explicitly flag which parts are official public information vs. inference or civilian observation.
Phenomenon / location:
Angle of interest:
Please provide:
1. Historical context (development timeline).
2. Institutional causes (how ordinances, planning policy, land leases etc. shaped the status quo).
3. Comparison with other Hong Kong districts or cases (if applicable).
4. Flag which information is backed by official documents / archives vs. oral or speculative.
Plan content:
Specific concern:
Please assess:
1. Reasonableness of the plan design (against usage forecasts, surrounding amenities).
2. Whether similar past infrastructure projects saw comparable controversy, and how they ended.
3. Potential execution risks or contentious points.
4. Which assessments need updated official data to confirm (e.g. latest usage statistics).
點擊展開 ▾Click to expand ▾事實查核Fact Checking
主張拆解（Claim Decomposition）Claim Decomposition
用途Use將一個複合主張拆解成可個別查核嘅子主張將一個複合主張拆解成可個別查核嘅子主張
主張：
請：
1. 將呢個主張拆解成獨立、可個別驗證嘅子主張
2. 對每個子主張，標明佢屬於：可查證嘅事實／統計數字 vs. 詮釋性判斷 vs. 預測
3. 標示邊個子主張最關鍵（推翻咗會令成個主張站唔住腳）
Claim:
Please:
1. Decompose the claim into independent, individually verifiable sub-claims.
2. For each sub-claim, flag whether it is: a verifiable fact / statistic vs. an interpretive judgement vs. a prediction.
3. Flag which sub-claim is most critical (overturning it collapses the whole claim).
Fact claim:
Existing sources:
Please provide:
1. Suggested independent source types to check (official statistics, primary reports, academic literature).
2. If sources disagree, how to judge which is more credible (recency, source nature, methodology transparency).
3. Current credibility rating of this claim given available info (high / medium / low / undetermined).
Misinformation:
Spread context:
Please analyse:
1. Which parts are factual errors, which are exaggerations or taken out of context.
2. Why this kind of info is easily believed and spread (emotional appeal, confirmation bias, source disguised as authority).
3. Suggested correction angle (how to correct without triggering a backfire effect that entrenches the false belief).
點擊展開 ▾Click to expand ▾Meta / 自我審查Meta / Self-Critique
Existing prompt:
Actual output problem:
Desired effect:
Please diagnose:
1. Which part of this prompt caused the problem (unclear instruction / missing example / undefined output format / insufficient role setup).
2. Explain the diagnosis point by point.
3. Provide an improved version.
4. Suggest a simple way to verify the improved version actually fixes the problem.
點擊展開 ▾Click to expand ▾Meta / 自我審查Meta / Self-Critique
Task description:
Desired output format:
Known constraints or boundaries:
Please design a structured prompt with:
1. Role setup.
2. Task instruction (broken into clear steps).
3. Output format spec.
4. Situations where the model should proactively flag uncertainty or request clarification.
5. Suggested examples to include (if applicable).
Below is a full analysis:
Reporting purpose:
Please compress it into an Executive Summary, requiring:
1. No more than 150 words (or within 5 bullets).
2. Lead with the conclusion, no narrative build-up.
3. Every bullet backed by data.
4. End with one line: "recommended decision" or "resources / approval needed".
Communication purpose:
Target department:
Background:
Expected action from them:
Please draft a formal email, requiring:
1. Open by stating the purpose concisely.
2. Give enough background, but avoid over-length.
3. Clearly list the specific action and deadline expected from them.
4. Tone: professional, objective, collaborative.
Core conclusion:
Analysis method:
Known limitations:
Please play a rigorous challenger and pose the 5 most likely sharp questions (covering methodology, data quality, conclusion robustness), and for each provide an honest, evidence-based response suggestion. If some question cannot be fully answered by current data, state it explicitly — do not force an answer.
點擊展開 ▾Click to expand ▾通用工具Utility Prompts
摘要（Summarise）Summarise
用途Use
請將以下內容摘要成 內，保留核心論點同關鍵數據，去除例子同鋪陳：
Please summarise the following content within , keeping the core argument and key data, removing examples and padding:
點擊展開 ▾Click to expand ▾通用工具Utility Prompts
比較（Compare）Compare
用途Use
請比較以下 項：
比較維度：
請用對照表呈現，並喺表後用一段講清楚整體結論。
Please compare the following items:
Comparison dimensions:
Present as a side-by-side table, then add one paragraph stating the overall conclusion.
From the content below, extract all (e.g. dates, numbers, names, action items):
Output as a structured list, and flag the original location of each item (if applicable).
點擊展開 ▾Click to expand ▾通用工具Utility Prompts
分類（Classify）Classify
用途Use
請將以下項目按 分類：
請講清楚每項嘅分類依據，並標示邊啲項目屬於邊界模糊、可能歸入多於一類。
Please classify the following items by :
State the basis for each item's classification, and flag items on blurry boundaries that may fit more than one category.
Please convert the following content from to :
Preserve the original meaning; do not add or drop substantive content. If any content cannot be fully preserved in conversion, flag it explicitly.
點擊展開 ▾Click to expand ▾文化/文本/電影/書籍Culture / Media / Books
I want to compare the following works:
Comparison dimensions:
Please analyse in this structure:
1. Each work's performance on each dimension (present side-by-side).
2. Commonalities and key differences, with the likely creative intent or temporal context behind them.
3. Which work handles which dimension more elegantly, with clear judgement basis (not just "better" — explain concretely why).
4. Flag which evaluations are objective, observable craft analysis vs. subjective taste judgements.
點擊展開 ▾Click to expand ▾文化/文本/電影/書籍Culture / Media / Books
文化現象／流行文化研究Cultural Phenomenon / Pop Culture Study
Phenomenon description:
Angle of interest:
Please provide:
1. The phenomenon's temporal context (how it rose, key turning points).
2. Analyse the driving factors from the .
3. Similarities/differences with past comparable cases (if applicable, give comparison examples).
4. Explicitly flag which analysis is based on verifiable facts / reports vs. inference or interpretation.
點擊展開 ▾Click to expand ▾文化/文本/電影/書籍Culture / Media / Books
Work name:
Analysis focus:
Please analyse:
1. Musical style features (rhythm, harmony, arrangement elements), and which musical tradition these elements inherit from.
2. Where this work sits in the artist's/band's creative arc.
3. Cultural positioning: how this work embodies / responds to urban nostalgia or a specific aesthetic identity.
4. If specific lyrics are involved, only paraphrase the theme in your own words — do not quote lyrics line by line.
點擊展開 ▾Click to expand ▾文化/文本/電影/書籍Culture / Media / Books
Book title:
Subject category:
Reading purpose:
Please provide a reading-notes framework:
1. Core argument (summarise in your own words, don't recount the original paragraph by paragraph).
2. Argumentation method and key evidence.
3. Possible links between this book and your other subject categories (e.g. Hong Kong urban systems, interdisciplinary thinking).
4. Critical questions: any obvious limitations, timeliness issues, or insufficiently verified assumptions in this book's argument.
5. Suggested further reading directions.
點擊展開 ▾Click to expand ▾文化/文本/電影/書籍Culture / Media / Books
Theme:
Related works / texts list:
Please organise:
1. How each work handles this theme (technique, angle).
2. Whether the treatment of this theme shows clear evolution over time.
3. Which treatments are industry conventions, which are individual creators' unique techniques.
4. Suggest a simple classification framework to help you categorise newly encountered works later.