On this page
THE SHORT ANSWER
A useful question says what happened, what matters to you, what you do not know and what you want from this exchange. A short account with one real constraint usually gives the Agent more to work with than a request to analyze your entire life.
- Describe observable events and label assumptions about motives separately.
- Include a relevant limit: time, budget, missing information or something you have already tried.
- Ask for a concrete result, then correct a specific omission in the next reply.
Will I regret changing jobs?
Do they care about me?
Am I suited to design?
Find my exact birth time.
The cards select work, relationships, learning and birth information, pairing each with an initial question such as “Will I regret changing jobs?” These are starting questions, not complete requests. The five-row table below supplies the fuller rewrites and adds a research example.
Start with what happened
Your first message does not need to be polished. Describe an event, then where you feel stuck. “Work has been frustrating” might become: “Requirements keep changing at short notice. I am unsure whether to keep adapting or discuss boundaries with the project lead.”
If finding the words is difficult, ask: “Help me narrow this down. Ask one question at a time, with a few options and space for my own answer.” The choices are suggestions you can change, not a test you have to pass.
Try this starting point
“I am dealing with [situation]. What matters most is [priority or concern]. I know [fact], but I am unsure about [missing information]. Please help me [one task].” Fill in only the parts you know.
For example: “I want to talk to my housemate about our different schedules. I need more quiet at night. They often come home late, but we have not discussed it properly. Help me find an opening that does not sound accusatory.” A conversation can help with this directly; you do not need a card draw first.
Five questions you can make more useful
| Situation | First attempt | A more workable request |
|---|---|---|
| Work | Will I regret changing jobs? | I value regular hours and know the salary, but not the travel schedule. List three things to ask before accepting. |
| Relationships | Do they care about me? | They canceled our last two meetings. I feel disappointed but do not know why. Help me ask without assuming a motive. |
| Learning | Am I suited to design? | I can try design for two hours a week. Suggest a two-week exercise and what to reflect on afterward. |
| Birth information | Find my exact birth time. | My family remembers between six and eight in the morning. Explain which fields may differ and what records I could check. |
| Research | Which solar-time method is most accurate? | Explain longitude correction and the equation of time. Separate astronomical definitions from Wenbu’s conventions and school differences. |
On a small screen, scroll the table sideways.
Keep the concern; make the question answerable
Focus on something you can observe, check or act on, while keeping the issue that matters to you.
- Work: “Will I definitely succeed?” → “What do I need to find out before accepting this opportunity?”
- Relationships: “What are they really thinking?” → “What have I observed, what am I assuming, and how could I ask?”
- Self-reflection: “Am I just unlucky?” → “Help me examine a recurring difficulty using a recent example.”
- Learning: “Which school is most accurate?” → “Which calculation rules differ, and what sources describe them?”
Give the next reply a clear direction
If you want to use tarot or I Ching, add: “Use the symbols as prompts, and keep track of what I still need to check in real life.” For research, name the concept, the claim you have encountered and the comparison you want.
Then follow up on a specific point: “Your second point is closest to my concern. Can you give an example?” Or: “That suggestion is beyond my budget. What could I try in two hours a week?” You can start with one sentence now and refine it through the conversation.
Worked example: discussing last-minute requests at work
You begin with “Am I in the wrong job?” Start by examining one recent situation before deciding what it means for your whole career.
Write the observation: “Twice this week, a new request arrived just before I finished work. Both times I stayed late.” Do not present disrespect as an established motive.
Add your priority and limit: “I want fewer late changes, but Friday’s delivery cannot slip. I have not asked which requests genuinely need same-day completion.”
Name the task: “Help me prepare three sentences to check priorities with the project lead. Do not decide whether I should leave.” The answer now has a specific job to do.
Revise a mismatch: “The second sentence is too formal for our chat messages. Make it shorter but keep the question about delivery priorities.” Remove any background the model invented.
Try: “This new request would take the evening. Which comes first: task A or request B? If B is needed tonight, can A move to tomorrow?” Record the reply before deciding on a follow-up.
A clear request cannot guarantee cooperation. It does let you check whether the answer actually prepares you for the conversation.
Key terms
- Observation
- Something you witnessed or have a record of, with its scope intact.
- Inference
- An explanation that needs evidence beyond the observation itself.
- Constraint
- A practical limit the answer should respect, such as time or missing data.
Common questions
Will more personal information make the answer more accurate?
Not necessarily. Share details that could change the advice. Names, addresses and other people’s contact information are usually unnecessary.
What if I cannot identify my goal?
Ask for one question at a time with two or three options and space for your own answer. You can remain unsure.
Does a clear prompt guarantee a correct answer?
No. It can clarify the task, but the model may still miss a constraint or misuse a source. Check important claims.
Sources and their scope
Supports how topic, goal, background and method form a draft. Work and relationship examples are original exercises.
Supports focused clarification, answer choices and rules against invented context or motives. These are product rules, not guarantees of every model response.
Traditional interpretations are not established predictions. See the source notes for context.
Read this guide with your agent
Get the complete text, diagram descriptions and sources in an open format. These files contain public knowledge only.
ADDITIONAL PRACTICE · 2026-11-09
Name the output first, then work backwards to the inputs
Turn a vague concern into a workable question: five rewrite exercises
This extension does not repeat the earlier explanation. It adds an exercise: write down your current question, rewrite it across five slots (fact, concern, unknown, constraint, request), then keep one action you can take this week. The five examples cover work, relationships, learning, birth information and source checking. A rewrite gives the answer a checkable job; it does not guarantee a correct conclusion, cooperation from others, complete records or a model that respects every limit.
Current en canonical article — preserve it and add a distinct useful exercise ↗Wenbu · asking a better question ↗Wenbu · agent integrations ↗
- The exercise aims at a checkable task, not a more accurate prediction.
- Each example keeps one action for the week and names what to observe.
- With uncertain birth data, ask which fields the uncertainty affects before asking for an exact time.
- Tarot and I Ching draws cannot be reproduced from the same question; record the procedure and the original result.
Four steps of one rewrite exercise
- 01Keep the original wording
Write the first vague sentence exactly as it came, unchanged, as a comparison point.
- 02Add facts and limits
Include one observable event and one real constraint, and label judgments about motives as inferences.
- 03Narrow to one request
Keep a single product, such as three sentences, a short list or a two-week exercise.
- 04Set a check
Note the action for this week, what to observe and when to look back; for draws, record the procedure and the original result.
Current en canonical article — preserve it and add a distinct useful exercise ↗Wenbu · asking a better question ↗Wenbu · reviewing a reading ↗
Why add an exercise instead of restating the advice
The existing article already explains what a useful question contains and gives a five-row rewrite table. The value of this extension is not another definition but a hands-on move: keep your original wording, fill the slots, then cut down to a single request. The point is to judge whether an answer did the job, not to believe one phrasing is luckier than another.
The exercise also keeps uncertainty visible. A rewrite reduces the chance of misunderstanding the task, but the existing article warns that a model may still miss a constraint, invent an explanation or misuse a source. Each example therefore carries a check: which sentence is an observation, which is an inference, and which step must be confirmed in real life. Better wording frames the question; it does not certify reality.
Current en canonical article — preserve it and add a distinct useful exercise ↗Wenbu · asking a better question ↗
Five slots: from one sentence to an executable request
Think of a rewrite as five slots: what happened (an observable event), what matters (a priority or concern), what is unknown (the information gap), what limits apply (time, budget, missing records), and what you want from this exchange (one concrete product). You need not fill every slot, but an event, one constraint and one request keep the answer from dissolving into general encouragement. Read it back and label judgments such as “they do not care” or “I am wrong for this field” as inferences.
The slots give the request a boundary. “Analyze my career” has none, so the answer can only spread out. “Two requests arrived late this week; I want fewer of them, Friday's delivery cannot slip, and I need three sentences to check priorities” has a product you can inspect. Rewriting is rarely about writing more; it usually means swapping an adjective for something that happened and a wish for something that can be delivered.
Current en canonical article — preserve it and add a distinct useful exercise ↗Wenbu · asking a better question ↗
Five before-and-after examples: work, relationships, learning, birth data, sources
Work: “Will I regret changing jobs?” becomes “I am choosing between two roles and value regular hours; I know the salary but not the travel schedule; list three things to ask before accepting, without deciding for me.” Relationships: “Do they care about me?” becomes “Our last two meetings were canceled at short notice; I feel disappointed and do not know why; help me ask about scheduling without assuming a motive.” Learning: “Am I suited to design?” becomes “I can spend two hours a week and do not yet know which branch I like; suggest a two-week exercise and what to review afterward.”
Birth data: “Find my exact birth time” becomes “My family remembers only between six and eight in the morning; first explain which fields this uncertainty affects and which original records I could check.” Sources: “Which solar-time method is most accurate?” becomes “Explain longitude correction and the equation of time, state the calculation conventions, and separate astronomical definitions from school conventions.” All five replace an unconfirmable outcome with something observable, checkable or actionable. None guarantees cooperation, and none creates records that do not exist.
Current en canonical article — preserve it and add a distinct useful exercise ↗Wenbu · asking a better question ↗Wenbu · agent integrations ↗
Random draws and calculations: what can be recomputed, what can only be recorded
If an exercise involves charts, separate two kinds of material. Calculations such as bazi or ziwei can be recomputed from the same date, time, timezone and conventions; Wenbu's integration page says calculation tools return structured results together with their conventions, and that ziwei does not automatically correct true solar time. When birth data is uncertain, the sensible request is to ask which fields the uncertainty affects, not to demand an exact time.
Tarot and I Ching draws are different: asking the same question again will not produce the same cards or hexagram. What can be reproduced is the record, not the draw. Note how many cards were used, whether reversals were allowed, and whether the I Ching cast was random or manual, then keep the result as it came. In review, compare what you did and observed rather than treating the result as a validated prediction. The integration page describes a full-deck draw without replacement and random or manual casting; these are procedure notes, not accuracy claims.
| Situation | First attempt | Rewritten request | What to check |
|---|---|---|---|
| Work | Will I regret changing jobs? | I am choosing between two roles and value regular hours; I know the salary but not the travel schedule; list three things to ask before accepting, without deciding for me. | Whether those three questions were actually asked |
| Relationships | Do they care about me? | Our last two meetings were canceled at short notice; I feel disappointed and do not know why; help me ask about scheduling without assuming a motive. | How the other person actually responded |
| Learning | Am I suited to design? | I can spend two hours a week and do not yet know which branch I like; suggest a two-week exercise and what to review afterward. | Whether the exercise was finished and how long it took |
| Birth data | Find my exact birth time. | My family remembers only between six and eight in the morning; first explain which fields this uncertainty affects and which original records I could check. | Whether a record that can be checked was found |
| Sources | Which solar-time method is most accurate? | Explain longitude correction and the equation of time, state the calculation conventions, and separate astronomical definitions from school conventions. | Whether the cited definitions and conventions match |
Current en canonical article — preserve it and add a distinct useful exercise ↗Wenbu · asking a better question ↗Wenbu · agent integrations ↗
Hypothetical exercise: turning “Am I wrong for this job?” into a checkable question
This is a hypothetical practice example. The person and events are fictional; it is not a user record and not a computed birth chart. It shows how one vague worry becomes a single piece of preparation for a conversation.
- Copy the original question unchanged: “Am I wrong for this job?” Keep it as the comparison point.
- Add one observable event: “Twice this week a new request arrived just before I finished, and both times I stayed late.” Do not write “the lead disrespects me,” because that is an inference.
- Add the concern and the limit: “I want fewer late changes, but Friday's delivery cannot slip; I have not asked which requests truly need same-day completion.”
- Narrow the request: “Help me prepare three sentences to check priorities; do not decide whether I should leave.” Ask for one product only.
- Set the check: talk to the lead once this week and record the reply. If the conversation does not happen, record why rather than judging the question by the outcome.
- If you want a tarot or I Ching draw as an additional prompt, add: “Use the symbols as prompts and list what I still need to check in real life.” Keep the procedure and result as recorded, not as reproducible evidence.
A week later, review whether the three sentences were used, how the lead responded and whether late requests changed. The rewrite gave the conversation checkable preparation; it does not guarantee cooperation and cannot prove whether the job suits you, which still needs more real-world information.
Current en canonical article — preserve it and add a distinct useful exercise ↗Wenbu · asking a better question ↗Wenbu · reviewing a reading ↗
Questions you may have
Will rewriting my question make the answer accurate?
Not necessarily. The existing article says a clear input reduces the chance of misunderstanding the task, but a model may still miss a constraint or misuse a source. A rewrite gives the answer a checkable job; you still verify important facts yourself.
Wenbu · asking a better question ↗Current en canonical article — preserve it and add a distinct useful exercise ↗
Do I have to fill all five slots in every example?
No. Keep at least one observable event, one real constraint and one concrete request. Slots you cannot fill can say “not yet known,” which is useful information and keeps uncertainty in the answer.
Current en canonical article — preserve it and add a distinct useful exercise ↗
How should I ask when the birth time is uncertain?
Ask first which fields the uncertainty affects and which original records you could check, rather than demanding an exact time. Calculation results can be recomputed from the same inputs and conventions, but missing records do not appear because the question improved.
Wenbu · agent integrations ↗Current en canonical article — preserve it and add a distinct useful exercise ↗
Can a tarot or I Ching result be reproduced?
A random draw cannot be reproduced from the same question. What can be reproduced is the record: how many cards, whether reversals were allowed, random or manual casting, and the original result. In review, compare what you did and observed, not a validated prediction.
Terms in this guide
- Observable event
- Something you witnessed or have a record of, kept with its time and scope, such as two meetings canceled at short notice.
Current en canonical article — preserve it and add a distinct useful exercise ↗
- Inference
- An explanation offered for an event, such as “they do not care,” which needs evidence beyond the event itself.
Current en canonical article — preserve it and add a distinct useful exercise ↗
- Constraint
- A practical limit the answer must respect, such as two hours a week, a Friday delivery that cannot slip, or a birth time known only as a range.
Current en canonical article — preserve it and add a distinct useful exercise ↗
- Calculation conventions
- The settings a calculation tool reports with its result, such as timezone, day boundary and whether true solar time is corrected; the same inputs and conventions can be recomputed.
See the claim and source notes
The existing article asks readers to separate observable events from assumptions about motives and to include one real constraint.
Describe observable events and label assumptions about motives separately.
Current en canonical article — preserve it and add a distinct useful exercise ↗
The existing article states that a clear prompt does not guarantee a correct answer.
Does a clear prompt guarantee a correct answer? No.
Wenbu's integration page says MCP returns structured results and calculation conventions for the host model to interpret.
返回结构化结果与计算约定
Sources and review scope
This AI-assisted Wenbu article passed automated structure checks and a separate editorial model pass. It is not independent expert review or a paid-account product test.
- Current zh canonical article — preserve it and add a distinct useful exercise ↗Existing article, not a new external source. Avoid repeating its explanation. · 2026-10-20 · Fetched text
- Current en canonical article — preserve it and add a distinct useful exercise ↗Existing article, not a new external source. Avoid repeating its explanation. · 2026-10-20 · Fetched text
- Wenbu · asking a better question ↗Existing editorial practice · 2026-10-20 · Fetched text
- Wenbu · reviewing a reading ↗Existing editorial practice · 2026-10-20 · Fetched text
- Wenbu · agent integrations ↗Current first-party product documentation · 2026-10-20 · Fetched text
Write your first sentence now
Copy the question you most recently wanted to ask, add one observable event and one action you can take this week; the question does not need to be perfect first.
Write your first sentence now ↗If you can already rewrite a question across five slots, the next step is not prettier wording but reversing the order: first write what you want to end up with, then ask what inputs that product needs. The worksheet has three lines: output (one checkable thing, such as three sentences, a comparison table or an opening line), inputs (the dates, limits and events the product actually requires), and tool (conversation, a calculation tool or a library search). Then run a reverse check: does this output depend on external facts you do not have? If so, change the request to “list what I need to verify” instead of letting the model fill the gap. Output-first gives the answer an acceptance test; it does not guarantee a correct conclusion or anyone's cooperation.
Current complete en canonical article ↗Wenbu · asking a better question ↗Wenbu · agent integrations ↗
- Write the output first, work backwards to inputs, and choose the tool last; the reverse order tends to produce an unbounded wish.
- The output must be checkable: three sentences, a table, an opening line, not “help me feel clearer.”
- The reverse check asks one question: does this output depend on external facts I do not have?
- Output-first does not improve accuracy; it gives the answer an acceptance test.
Three steps of the output-first worksheet plus one reverse check
- 01Write the output
One noun-like product only, such as three sentences, a comparison table or an opening line, plus what counts as done.
- 02Work back to inputs
List only what the product needs: one event that happened, one real constraint and one known gap.
- 03Choose the tool
Conversation, a calculation tool or a library search, depending on what the output needs rather than what sounds strongest.
- 04Reverse check
If the output depends on external facts you lack, rewrite it as “list what needs verifying and how.”
Current complete en canonical article ↗Wenbu · agent integrations ↗
Why start from the output, not the feeling
The five-slot rewrite starts from what happened, which is a good way to turn a complaint into a complete request. But many people fill all five slots and still write “help me analyze this,” because their attention is still on the feeling. Output-first reverses the order: decide what you want to end up with, then go back for the inputs. Once the output is concrete, “analyze this” has nowhere to hide, because you can either point to the product or you cannot.
This order has a side effect: it exposes when you have no inputs at all. You may want the model to judge whether to move cities, then notice you have no figures on rent, commuting or family arrangements. The right output in that case is not a verdict but a list of what still needs checking. Admitting a missing input is more honest than letting the model supply a premise for you.
Current complete en canonical article ↗Wenbu · asking a better question ↗
A three-line worksheet: output, inputs, tool
Line one is the output: exactly one, written as a noun, such as three sentences to check priorities, a two-option comparison table, or an opening line that does not assume a motive. Line two is the inputs: only what the product genuinely needs, namely one event that happened, one real constraint and one known gap. Line three is the tool, and only now: plain conversation if that suffices, a calculation tool if a chart must be recomputed, a library search if sources are involved, noting that the search covers only its own catalogue.
The order cannot be shuffled. People who pick the tool first end up asking which tool is most accurate, but suitability depends on the output. A calculation tool cannot write an opening line, and speculation in a chat should not stand in for stated calculation conventions when a chart is being recomputed. Read the worksheet back: if the output line still says “understand fully” or “analyze deeply,” it is not narrow enough, so rewrite line one.
Current complete en canonical article ↗Wenbu · agent integrations ↗
The reverse check: does the output depend on external facts?
The last step is a reverse check. Ask of each item: for this output to hold, does it depend on facts I do not have and cannot obtain from the conversation, such as another person's private thoughts, unpublished shift schedules or a relative's memory of a detail? If so, change the output to “list what needs verifying and how,” rather than demanding a definite answer. The output stays checkable; what gets checked shifts from the conclusion to whether the list is complete.
The reverse check must also separate three kinds of tool. Wenbu's tarot draw is a documented random procedure, so asking the same question again cannot reproduce the same cards on demand; a coincidence is still possible, so the output should be “keep this draw's procedure and original result.” I Ching casting can be random or manual: a random cast likewise cannot be reproduced on demand, while a manual entry is your own six line values and can be repeated exactly, so record the six numbers and the changing lines. Calculated results differ again: the same inputs and conventions can be recomputed, so the output can be “recompute with the same inputs and compare the stated conventions.” Name which kind you are dealing with in the output line.
Wenbu · agent integrations ↗Wenbu · reviewing a reading ↗Wenbu · asking a better question ↗
| Concern | Output phrasing | Inputs needed | Reverse check |
|---|---|---|---|
| Whether to accept a new role | A list of three things to ask before accepting | Known salary, unknown travel schedule, regular hours as the priority | Travel frequency is an external fact, so the list stays a set of questions rather than a verdict |
| Talking to a housemate about schedules | An opening line that does not accuse | Late arrivals are observable; quiet nights are the priority | Whether they cooperate cannot be known, so the output stops at “opening line prepared” |
| Uncertain birth time | A list of affected fields and records to check | The remembered range, the known date and timezone | Missing records do not appear because the question improved, so the output cannot be “an exact time” |
Current complete en canonical article ↗Wenbu · agent integrations ↗
Hypothetical worksheet: turning “should I sign up for this course?” into a checkable output
This is a hypothetical practice example. The person and events are fictional; it is not a user record and not a computed birth chart. It shows how the output-first worksheet turns hesitation into something that can be accepted or rejected.
- Copy the original question unchanged: “Should I sign up for this course?” Keep it as the comparison point.
- Write the output: a list of three questions to ask before enrolling, each one askable in a single enquiry.
- Work back to inputs: the fee is known; two hours a week are available; the workload and the make-up policy are unknown.
- Choose the tool: plain conversation suffices; no calculation tool or draw is needed.
- Reverse check: workload and make-up policy are external facts held by the provider, so the output stays a set of questions and the model must not fill them in.
- Set the check: send one enquiry this week and record the answers to the three questions; if there is no reply, record why.
A week later, review whether the three questions were asked, how the provider answered and whether the answers changed the enrollment decision. The worksheet gave the enquiry checkable preparation, but it does not guarantee the course suits you or that the provider answers accurately; that still needs more real-world information.
Current complete en canonical article ↗Wenbu · asking a better question ↗Wenbu · reviewing a reading ↗
Questions you may have
Does output-first conflict with the five-slot rewrite?
No; they run in different directions. The five-slot rewrite starts from an event and completes the request; output-first fixes what you want to end up with and then looks back for events and limits. Readers already comfortable with the slots can use output-first as a finishing pass so the request does not drift back into “analyze this.”
Is “a list” too vague as an output?
Yes. Say how many items and what each must answer, for example “three things to ask before accepting, each askable in one call.” A vague output cannot be accepted or rejected, and the answer slides back into generalities.
Current complete en canonical article ↗Wenbu · asking a better question ↗
If the reverse check finds external facts, should I drop the question?
No; keep it but change the output to “list what needs verifying and how.” The answer remains checkable, with the check moving from the conclusion to whether the list is complete. Missing records do not appear because the question improved.
Wenbu · agent integrations ↗Current complete en canonical article ↗
What should the output be when tarot or I Ching is involved?
For tarot, write “keep this draw's procedure and original result,” such as how many cards and whether reversals were allowed. For I Ching, note whether the cast was random or manual: a random cast cannot be reproduced on demand and a coincidence is still possible, while a manual entry's six line values and changing lines can be recorded again exactly. The record is reproducible, not the random draw.
Terms in this guide
- Output
- The one checkable thing you want from this exchange, such as three sentences, a comparison table or an opening line, with a clear test of whether it is done.
- Reverse check
- Asking of each output whether it depends on external facts you lack; if so, rewrite it as a list of things to verify.
Current complete en canonical article ↗Wenbu · agent integrations ↗
- External fact
- Information that does not come from the conversation and needs a real record or another person's answer, such as shift schedules, provider policies or someone's private thoughts.
See the claim and source notes
The existing article asks readers to separate observable events from assumptions about motives and to include one real constraint.
Describe observable events and label assumptions about motives separately.
The existing article states that a clear prompt does not guarantee a correct answer.
Does a clear prompt guarantee a correct answer? No.
Wenbu's integration page says MCP returns structured results and calculation conventions for the host model to interpret.
返回结构化结果与计算约定
Sources and review scope
This AI-assisted Wenbu article passed automated structure checks and a separate editorial model pass. It is not independent expert review or a paid-account product test.
- Current complete zh canonical article ↗Current public original and earlier additions; add substantial coverage without repeating or replacing them. · 2026-11-09 · Fetched text
- Current complete en canonical article ↗Current public original and earlier additions; add substantial coverage without repeating or replacing them. · 2026-11-09 · Fetched text
- Wenbu · asking a better question ↗Existing editorial practice · 2026-11-09 · Read excerpt
- Wenbu · reviewing a reading ↗Existing editorial practice · 2026-11-09 · Read excerpt
- Wenbu · agent integrations ↗Current first-party product documentation · 2026-11-09 · Fetched text
- Wenbu · versioned public capabilities ↗First-party current tool, convention and privacy contract; not external endorsement or predictive validation · 2026-11-09 · Fetched text
Write your output line
Set aside the question you most recently wanted to ask and write the one checkable product you want first, then work backwards to its inputs; if you cannot name the inputs, turn it into a list of things to verify.
Write your output line ↗