Five short exercises: Context, Role, Action, Format, Trust. Write and discuss here. You build in F3.
01 · The frame
Five slots in every good prompt
Miss a slot and the agent guesses. Click each one, then try the warm-up.
C · Context · Exercise 1
Warm-up · 2 minutes with your team
Which slots are missing from each prompt? Click the blurred text to check.
Prompt A
"Summarize this document"
Missing:R, F and T. No role, no format or length, and nothing forcing it to stay inside the document — so you get a different summary every time you ask.(click to reveal)
Prompt B
"Tell me everything about this vendor proposal and whether we should go with them"
Missing:A and T mostly. "Everything" is not an Action, and asking "should we" with no criteria invites the agent to invent its own. No Trust slot means it will fill gaps in the proposal with assumptions instead of flagging them.(click to reveal)
Prompt C
"Write me an email to the team about the budget"
Missing:C, R and F. Which budget, and what actually happened to it? Who is writing, and to whom? And what is the email for — to inform, to request, or to escalate? All three change the draft completely.(click to reveal)
Notice that Trust is missing from all three.
How to use this page
Exercise 1 covers C and A, Exercise 2 is R, Exercise 3 is F, Exercise 4 is T, and Exercise 5 puts all five together. Work through them inside F2, on paper or in a doc, at the pace of the room. Nothing here needs Amazon Quick open.
The prompt you finish with has the same shape as a real Chat Agent brief. F3 supplies a ready-made one so the whole room builds in step, but you can paste yours in place of it, so keep it.
The failure to watch for: a prompt with missing slots rarely gives obvious rubbish. It gives something fluent, confident and quietly wrong, such as an answer that skips a rule nobody checked.
Exercise 1 · Context + Action
C ContextA Action
Name the document and the deliverable
Give the agent the actual document, and ask for a named output instead of "what do you think?".
1Try it
Scenario: a vendor proposal for medical equipment maintenance lands before the next procurement committee. Rewrite this prompt on paper, filling C and A first.
BEFORE — vague prompt
Look at this vendor proposal and tell me what you think.
2Look for
C: which document, and what must it be compared against?
A: a named artefact plus a judgement, not "review this".
A named deliverable beats "tell me what you think".
3 · Model answer: one strong prompt
AFTER — powerful prompt
You are a Senior Procurement Analyst with 10 years of experience in healthcare vendor assessment.
Read the attached vendor proposal and produce:
1. **Summary Table** with columns:
| Vendor Name | Service Offered | Contract Duration | Annual Cost (SGD) | SLA Uptime % | Key Risks |
2. **Compliance Checklist** — check these against our requirements:
- [ ] ISO 13485 certified (medical devices)
- [ ] PDPA compliant data handling
- [ ] 24/7 emergency response capability
- [ ] Local service team in Singapore
- [ ] References from healthcare sector
3. **Recommendation:** PROCEED / REVIEW FURTHER / REJECT
- One paragraph justification (max 100 words)
- List any missing information that would change the recommendation
Rules:
- Only use information from the attached document
- If any checklist item is not mentioned, mark as "NOT PROVIDED — verify with vendor"
- All costs in SGD
- Do not recommend PROCEED if any compliance item is missing
- Flag any cost that seems unusually low or high compared to typical healthcare maintenance contracts
Want to run it? Optional, after the session. Add a sample document to a Space, open a chat on that Space with Web search off (click + in the chat box), paste your prompt and compare it with the vague one.
Exercise 2 · Role
R Role
A persona sets the expertise and tone
Five parts make a persona that behaves like a specialist, not a generic assistant.
The formula, with an example
TitleSenior Finance Manager
Experiencewith 12 years
Specialtyin healthcare budgeting
Characteristicdetail-oriented and conservative
Behaviourwho always cites source data
1Try it
Write a persona with the formula for each scenario, then read a couple aloud at your table.
Scenario
Your persona (fill in)
A. Reviewing staff leave applications for HR
Try: "HR Operations Specialist with 8 years in healthcare workforce management..."
B. Drafting a memo to department heads about budget cuts
Try: "Chief Financial Officer with 15 years in public healthcare finance..."
C. Summarizing patient feedback surveys (non-clinical)
Try: "Patient Experience Manager with 10 years in hospital quality improvement..."
D. Checking a new MOH circular against current policies
Try: "Compliance Officer with 12 years in healthcare regulatory affairs..."
2Look for
Which persona would notice something a general assistant would miss? The behaviour clause is usually why.
In Quick this goes in the agent's persona instructions.
3 · Model answer: a persona in action
EXAMPLE — persona in action on Amazon Quick
You are a Senior Administrative Officer with 10 years of experience in hospital operations management. You are methodical, detail-oriented, and always prioritize patient safety considerations even in administrative decisions. You write in a professional but accessible tone suitable for department heads who are clinicians, not administrators.
Read the attached meeting minutes and produce:
1. A bullet-point summary of key decisions (max 5 bullets)
2. An action item table: | Action | Owner | Deadline | Priority (High/Medium/Low) |
3. Any items that may impact patient services — flag these with "WARNING:"
Only extract information explicitly stated in the document. If a deadline is not mentioned, write "TBD — follow up required".
Without a persona the reply is generic. With one, it takes on the vocabulary and judgement of that role: a "Senior Finance Manager" flags budget risks a generic reply would miss.
Exercise 3 · Format
F Format
Name the shape you want back
Sections, table columns and a word limit give you the same shape every time.
Monday's run
Next month's run
Two runs, two different shapes.
1Try it
Scenario: a monthly operations report for the Hospital Director. Rewrite this prompt so the structure is fixed.
UNSTRUCTURED
Create a monthly report from this spreadsheet data.
2Look for
Would you forward the result to your director without editing it? That question is the whole F slot.
Same shape every run means nothing to re-check.
3 · Model answer: a structured prompt
STRUCTURED — consistent output every time
You are a Hospital Operations Analyst preparing the monthly report for the Hospital Director.
From the attached spreadsheet, generate a report with EXACTLY this structure:
## Executive Summary (3 sentences max)
[Overall performance statement, key highlight, key concern]
## KPI Dashboard
| Metric | This Month | Last Month | Target | Status |
|--------|-----------|------------|--------|--------|
[Fill from data — use ON TARGET, WATCH (within 10% of target) or MISSED (more than 10% off target)]
## Highlights (max 3 bullets)
- [Positive achievement with specific number]
- [Positive achievement with specific number]
- [Positive achievement with specific number]
## Concerns Requiring Attention (max 3 bullets)
- [Issue + impact + suggested action]
- [Issue + impact + suggested action]
- [Issue + impact + suggested action]
## Next Month Focus Areas
1. [Priority action]
2. [Priority action]
3. [Priority action]
Rules:
- All percentages to 1 decimal place
- All costs in SGD with thousands separator
- If data is missing for any metric, write "Data not available — follow up with [department]"
- Keep total report under 400 words
- Tone: Professional, factual, no jargon
More structures for healthcare admin
Use case
Best output format
Why
Vendor comparison
Comparison table + recommendation
Side-by-side makes differences obvious
Policy compliance
Checklist with pass / fail / unclear status
Quick scan for gaps
Meeting minutes
Decisions + action items table
Actionable, easy to track
Budget analysis
Variance table + narrative
Numbers need context
Email drafting
Subject line + body + sign-off
Ready-to-send format
Risk assessment
Risk register (impact × likelihood)
Standard risk format
Want to run both? Optional, after the session: try them on Amazon Quick against any spreadsheet and see how far apart they land.
Exercise 4 · Trust
T Trust
Make the answer defensible
Trust has three habits: cite the section, say "not covered", name who decides. Pick a question to see them.
Question put to a leave-guideline assistant (GL-HR-014)
For each agent, write at least 3 "Must NEVER" rules and 2 "Escalate when" triggers. Think about what could go wrong in a healthcare setting.
Agent A: Patient Feedback Summarizer
Reads patient satisfaction surveys and writes a summary for the quality improvement team.
Must NEVER:
Include any patient names, NRIC, or identifiable information in the summary
Make clinical judgments about care quality
[Your turn — add more]
Escalate when:
Feedback mentions patient safety concerns or near-misses
[Your turn — add more]
Agent B: Purchase Order Validator
Checks purchase orders against approved budget, vendor list and procurement policy before routing for approval.
Must NEVER:
[Your turn — think about financial risks]
Escalate when:
[Your turn — what thresholds matter?]
Agent C: Staff Communication Drafter
Drafts internal emails and memos for department heads from bullet-point instructions.
Must NEVER:
[Your turn — what could go wrong with automated emails?]
Escalate when:
[Your turn — when should a human review before sending?]
2Look for
Every "escalate" rule names a person or role who takes over, not just "a human".
What an agent must not do matters as much as what it does.
3 · Model answer: guardrails for Agent B
GUARDRAILS — Purchase Order Validator
## You Must NEVER
- Approve or recommend approval of any purchase order (you only validate — humans approve)
- Process POs from vendors not on the approved vendor list
- Accept POs without a valid budget code and cost center
- Override procurement policy limits regardless of justification given
- Disclose budget remaining balances to requestors (only to finance team)
## Escalate to Human When
- PO amount exceeds $25,000 SGD (requires Director approval)
- Vendor is new (not on approved list) — route to Vendor Management
- Budget code shows insufficient remaining balance
- PO is for controlled items (pharmaceuticals, medical devices, IT equipment)
- Same vendor has 3+ POs in the same week (potential split-ordering to avoid thresholds)
- Requestor and approver are the same person
## Data Grounding Rules
- Only validate against the uploaded approved vendor list and budget file
- If a vendor is not found in the list, state "VENDOR NOT ON APPROVED LIST — requires Vendor Management review"
- If budget data is unavailable, state "BUDGET VERIFICATION PENDING — cannot validate without current balance"
Exercise 5 · All five slots
CRAFT all five
Write one complete prompt
Choose a scenario and write all five slots. Keep it: it has the shape of the Chat Agent brief you use in F3.
1Try it: choose a scenario
Pick a scenario to see what it needs in Quick.
2Tick each slot as you fill it
0 of 5 slots
If you cannot fill a slot, that is where the agent will guess.
3 · Check it before you leave this page
You do not need Amazon Quick for these. Read your prompt back and ask:
Does it produce the exact format you specified?
Given incomplete data, does it say "NOT PROVIDED"?
Asked something out of scope, does it refuse or escalate?
Is the tone right for your audience?
Would you trust the output enough to forward to your director?
Reference · Cheat sheet
The CRAFT template
Copy it, fill the brackets, and use it as your Chat Agent brief.
UNIVERSAL CRAFT TEMPLATE FOR AMAZON QUICK
# C — CONTEXT
Source of truth: [DOCUMENT NAME + VERSION] — held in this Space.
Reference material it must check against: [TEMPLATE / CHECKLIST / MATRIX].
This case: [THE SPECIFIC FACTS — dates, amounts, grades, quantities].
# R — ROLE
Act as a [TITLE] with [N] years at Synapxe, specialising in [SPECIALTY].
You are [CHARACTERISTIC], because [WHY IT MATTERS HERE — the behaviour clause].
# A — ACTION
[THE DELIVERABLE + THE JUDGEMENT. Name an artefact, not an activity.]
1. [Read / extract — what to pull from the document]
2. [Compare / assess — what to do with it]
3. [Produce — what comes out]
Do not simply summarise the source.
# F — FORMAT
[LENGTH]. Sections in this order:
1. [SECTION]
2. [SECTION] — table with columns: [COL] | [COL] | [COL] | Status
3. [SECTION]
Save as a Markdown (.md) file.
# T — TRUST
Answer ONLY from the documents in this Space.
Cite the clause, section or page number for every claim.
If the source does not address something, write "NOT COVERED IN [DOC]"
and route it to [ESCALATION PATH] — never fill the gap from general practice.
You must NEVER:
- Fabricate or assume data that isn't in the source
- [DOMAIN-SPECIFIC PROHIBITION]
- Include patient names, NRIC or identifiable health information
- Provide medical, legal or financial advice
Escalate to a human when:
- [THRESHOLD TRIGGER]
- [SAFETY OR EXCEPTION TRIGGER]
- You are unsure of the correct interpretation
All monetary values in SGD.
C goes in the Space. The exact T wording goes in a reference document.
Where each slot lives in Quick
Add the document to a Space once, not once per conversation. Quick rewrites persona instructions with generative AI, so wording that must survive exactly (the "NOT COVERED" phrase, the citation rule, the refusals) belongs in a reference document, which keeps its wording verbatim. Keep a short version of T in the persona instructions too. A is what you type in the chat each time.
Standard Trust rules for healthcare admin
Paste into the T slot of every Synapxe agent
- Never include patient names, NRIC numbers, or identifiable health information in outputs
- Never provide clinical advice, diagnoses, or treatment recommendations
- Never approve expenditures — only validate and recommend (humans approve)
- Always cite the source document and page/section for any claim
- State "NOT PROVIDED" for any information not found in the documents in the Space
- Escalate to a human when patient safety may be impacted
- All monetary values in SGD
- Follow PDPA guidelines — no personal data in summaries unless explicitly required
Power phrases
Instead of...
Say this...
Why it works
"Summarize this"
"Extract the top 5 key points as bullet points, max 20 words each"
Specific format + length
"Is this good?"
"Evaluate against these 5 criteria and rate each 1-5 with justification"
Measurable assessment
"Write an email"
"Draft a professional email to [audience] informing them of [topic]. Tone: [formal/friendly]. Max 150 words."
Audience, purpose, tone, length
"Check this document"
"Compare this document against [reference] and produce a compliance checklist marking each requirement pass, fail or unclear"
Clear comparison
"Help me with this"
"You are a [persona]. Read the attached [document type] and produce [specific output format]"
Role, input and output defined
"What do you think?"
"Based only on the data provided, list 3 strengths and 3 risks. For each risk, suggest one mitigation action."
Grounded, structured, actionable
Common mistakes
Don't
Ask open-ended questions ("what do you think?"): A unfilled
Leave the output shape to the agent: F unfilled
Skip the "must NEVER" rules: T unfilled
Assume it knows your situation: C unfilled
Use jargon the agent might read differently to you
Do
Walk the five slots before you run anything
Name exact table columns and a word limit (F)
Write the "must NEVER" and escalation rules (T)
Give it the words to use when something isn't covered
Test with edge cases: missing data, out-of-scope requests