This page provides a manual prompt template for Claude-style task instructions. XML-style tags can make source material and requested output easier to distinguish; this page does not run Claude. Fill the worksheet, copy the completed prompt into the relevant workflow, and review the actual result. The examples below are editorial drafts, not observed outputs.
Illustrative concept image made for this page · not a tested product output.
Claude Prompt Worksheet
Replace each bracket with your own information. Keep unknown facts out of the brief.
Prompt text
<task>[one concrete request]</task>
<context>[approved facts or source text]</context>
<constraints>[what must stay exact; what to flag]</constraints>
<output_format>[the structure you need]</output_format>
<review>[what a person must verify]</review>
Which brief fits your source?
Brief 1: Claude XML task worksheet. Brief 2: Travel-mug video outline example.
Selection rule: Use XML tags to separate instructions from source facts; use the mug example only with approved front and lid photos.
Source-to-instruction relationship: The context block supplies facts; constraints specify what stays exact; a well-formed tag cannot verify a claim by itself.
If the draft misses the brief
Visible signal: The outline adds temperature retention or price because the XML structure looks complete.
Next revision: Keep the missing fact flagged, and ask for evidence before writing that claim into a scene.
Keep the source and review together
Source to retain: the completed XML context with approved mug facts and source photos. After an actual run: save the exact prompt version, resulting file and a review note stating whether price and temperature claims stay unresolved until the owner supplies evidence. Until that run exists, leave the result and review fields unfilled. The Gallery cases below are separate published projects and cannot complete this record.
Separate Context with XML Tags
Use descriptive tags to separate source content from instructions. Keep each section short enough to inspect. State how to handle missing or conflicting facts rather than asking the model to fill them in.
Official Pexo Gallery · selected references
See these ideas in motion
Compare 2 separately published Pexo Gallery examples related to this page's visual or workflow choices.
Official Gallery examples with Prompts published on their source pages. Related drafting and text-review visuals only; neither shows Claude, XML tags or a tested Claude output. These videos are separate from this page's starter briefs, supplied assets and any named model. Review your own result before publishing.
Related Official Pexo Gallery example · separate published project
Inkly AI Writing Workflow Demo
An AI writing-service animation moves from glowing input bars to a phone document and a Draft/Edit/Ready menu. The writing document and Draft/Edit/Ready states provide a visual reference for reviewing a structured brief before reuse.
Prompt published with this Gallery video
Create a neon purple AI writing product demo that moves from an idea input into generated copy, document editing, and mobile previews. Use luminous…
View full prompt ▾Show less ▴
Create a neon purple AI writing product demo that moves from an idea input into generated copy, document editing, and mobile previews. Use luminous gradients, crisp panels, quick interface transitions, and readable product states.
Related Official Pexo Gallery example · separate published project
Apple Notes Checklist and Drawing Tools
A notes interface shows a Copy menu, a grocery checklist with visible checkmarks and a drawing-tool animation. The Notes copy menu and checklist make separate text items visible, a limited analogy for keeping task and source material inspectable.
Prompt published with this Gallery video
Create a concise Apple Notes product walkthrough. Show creating a new note, building a grocery checklist, and using handwriting, highlighting, and d…
View full prompt ▾Show less ▴
Create a concise Apple Notes product walkthrough. Show creating a new note, building a grocery checklist, and using handwriting, highlighting, and drawing tools with familiar mobile gestures, clean typography, and responsive interface motion.
Example 1: Task: turn approved notes into a short video outline.
Prompt text
<task>Draft a three-scene product video outline.</task>
<context>Approved notes: blue travel mug; screw lid visible in the supplied front and lid photos. No price, temperature claim or extra feature has been approved.</context>
<constraints>Keep the blue color and screw lid exact; flag missing price or performance claims rather than writing them.</constraints>
<output_format>Scene, visual, narration, fact to verify.</output_format>
Example 2: Task: summarize interview notes.
Prompt text
<task>Summarize the supplied interview notes.</task>
<context>Use the transcript I provide, preserving speaker attribution.</context>
<constraints>Do not invent quotations or motivations.</constraints>
<output_format>Five findings and two open questions.</output_format>
These are written example prompts; they are not screenshots or measured outputs.
From tagged input to a checkable outline
In the filled example above, <task> says what to make, <context> contains the only approved facts, <constraints> tells the model what to leave out, and <output_format> names the columns to inspect. A useful review pass asks: Does each scene point to one supplied photo? Does narration introduce a price or heat claim that was never provided? If it does, correct the specific scene and rerun that request with the source notes unchanged. This is an illustrative workflow, not a Claude result.
Check what the tags actually protect
Put approved source facts inside <context> and the requested action inside <task>; a sentence quoted from a customer email belongs in context even if it sounds like an instruction. For the outline example, compare each proposed section with the supplied facts and flag missing evidence instead of adding a plausible statistic. XML-style boundaries make review easier, but they do not verify the source or the answer.
Review and Iterate
Tags organize the text; they do not verify source truth or force a particular model behavior. Check the current documentation and output in your own Claude workflow.
Carry a Video Brief to Pexo
When the reviewed output is a video brief, bring it and authorized assets to Pexo's AI video generator. Review scenes, claims and captions before use.
YOUR NEXT STEP IN PEXO
Bring a reviewed brief into Pexo.
Use authorized assets, inspect the draft, and revise what does not match the brief.
Use only references you have permission to submit. Remove unnecessary personal or customer information. Before sending sensitive material to any service, review its current terms and privacy policy; this worksheet makes no claim about storage, deletion or account conditions.
Keep instructions outside source material
Use this brief when: Use XML-style tags when a Claude task contains approved facts, examples and output instructions that should not be confused.
Acceptance check: Check that each fact is inside the context block, each request is in the task block, and the output flags missing or conflicting information.
If it fails: If source text starts acting like an instruction, move it back into a labeled context block and retest.
Suppose a draft of the travel-mug outline says “keeps drinks hot for 12 hours,” while the approved context only shows a blue mug and screw lid. That sentence would be unsupported; it is a hypothetical failure example, not an observed Claude output. A targeted follow-up brief is:
Prompt text
Revise only the product-claim sentence in scene 2. The approved source establishes a blue travel mug and visible screw lid, but gives no temperature or duration performance. Remove the 12-hour claim. Keep the three-scene order and mark any other unsupported product claims for owner review.
Compare the revision with the approved notes again before using it as a video script. Clear XML tags can help separate facts from instructions, but cannot validate missing evidence.
Claude Prompt FAQ
Is XML required for every Claude prompt?
No. Use tags when they help separate instructions, source text and output.
Do XML tags verify the source?
No. Tags organize instructions and context; a person still needs to check claims, quotes and omissions against the source.