Free templates · Independent Claude learning
Claude Prompts for Work: 3 Templates with Review Checks
The best starting point is a work product you can inspect, not an impressive-sounding role. These three original templates turn ordinary workplace inputs into drafts with visible gaps. Each includes a check so that a polished response does not hide an invented fact.
Build one repeatable drafting workflow, check it against its inputs, and learn which prompting skill to practise next.
By ClaudeQuest · Updated
Go to the free templates ↓Copy, adapt, and check
No registration is required to read or copy these templates. Replace placeholders and check the result against your own input.
1. Turn meeting notes into an action register
Use anonymized notes. Example input: ‘Mira will draft the launch checklist. We discussed Friday but did not agree a deadline. Budget approval is still open.’
Turn the notes below into an action register for the project owner.
Return a table: Action | Explicit owner | Agreed deadline | Evidence from notes.
Use 'Not specified' when an owner or deadline is absent.
Separate decisions from suggestions. Put unresolved questions below the table.
Do not assign people, invent dates, or send messages.
<notes>
[Paste approved, anonymized notes here.]
</notes>How to check the result
Expected for the example: Mira owns the checklist; deadline is Not specified. Friday belongs in unresolved questions, not in the agreed-deadline field. Budget approval must not be described as complete.
2. Draft a weekly status update without invented progress
Example input: ‘Search filter merged; mobile check not run. Import issue reproduced, cause unknown. Need a test file from the owner.’
Draft a weekly update for a busy project lead using only the notes.
Use headings: Verified progress, Risks, Next actions, Decisions needed.
Keep it under 180 words. Preserve uncertainty and incomplete checks.
Do not turn 'planned', 'reproduced', or 'merged' into 'verified in production'.
For each next action, include the owner only if the notes name one.
<notes>
[Paste approved status notes here.]
</notes>How to check the result
The draft may say the filter was merged, but must retain the missing mobile check. The import cause remains unknown. A request for a test file is a dependency, not proof that the issue has been resolved.
3. Compare two proposals with a source trail
Example inputs: Proposal A states a four-week timeline and no support terms. Proposal B states six weeks and includes 30 days of support. Neither states a price.
Compare the two proposals for a project lead.
Criteria: delivery timeline, included support, stated price, unresolved assumptions.
Return: Criterion | Proposal A evidence | Proposal B evidence | Question to resolve.
Use 'Not stated' for missing information. Do not infer missing commercial terms.
After the table, list information needed before a decision; do not select a winner.
<proposal_a>[Approved text A]</proposal_a>
<proposal_b>[Approved text B]</proposal_b>How to check the result
Both prices must be Not stated. A's missing support terms do not prove support is excluded. The shorter timeline alone is insufficient to declare A the better purchase.
Why these prompts use explicit inputs and checks
Anthropic's prompting guidance recommends clear instructions, relevant examples, and distinct boundaries between instructions and supplied content. Here, descriptive tags separate task data from the request. Tags improve organization; they are not a security barrier, a truth guarantee, or a replacement for checking the result.
Choose the smallest useful task
Begin with a recurring draft that you already know how to review. A weekly update is a better first experiment than delegating an entire project. Record the time spent preparing inputs, correcting the draft, and completing the review. If preparation and correction take longer than writing it yourself, simplify the workflow before adding automation.
Test missing information on purpose
Try each template with one complete input, one missing fact, and one contradiction. For meeting notes, remove a deadline and check whether it stays unknown. For a status update, include a failed check and see whether the failure survives summarization. For proposals, give conflicting delivery dates and expect an unresolved question. These tests expose failure modes that a single polished example cannot.
Protect the human decision boundary
These are drafting prompts, not instructions to send an email, modify a record, approve a budget, or choose a supplier. Review the work before it leaves your workspace. Use data your organization permits, remove unnecessary personal information, and keep sensitive material out of public examples. A fluent draft can still omit a critical qualification.
Common mistakes and better alternatives
Weak approach: Summarize this and make it sound successful.
Improve it: Ask for verified progress and unresolved risks separately. Preserve failed or incomplete checks.
Weak approach: Fill in the missing owner and deadline.
Improve it: Keep unknown fields explicit and list the decision needed from the project owner.
Weak approach: The first output looks good, so automate it.
Improve it: Test missing and contradictory inputs, then measure correction effort before automating.
Questions before you start
Do these templates require the Claude API?
No API integration is required for the text exercises. They are drafting patterns, not automated workflows. Check your organization's approved tools and data-handling rules before using real work material.
Are the examples real customer data?
No. All examples on this page are fictional learning examples created by ClaudeQuest. They demonstrate review criteria, not measured productivity results.
What should I learn after copying a prompt?
Learn to identify missing context, specify a reviewable output, and test uncertainty. The free skills check gives feedback on a small sample; it does not certify workplace proficiency.
Sources and scope
Anthropic Claude Platform documentation: Prompting best practices. Technical source checked September 7, 2026. Templates, fictional inputs, and expected-output checks are original ClaudeQuest material.
ClaudeQuest is independent of Anthropic. These templates are not official exam material, certification, or a guarantee of output quality.
How we review learning material