You can use this approach for a simple product assembly, an app's first useful task, or a setup procedure at work. The guide should get a particular reader from a known starting point to one useful outcome. Its length depends on the task, the pictures needed to explain it, and the place where someone will read it.
Follow a worked example from brief to step card
The Pinefold example PDF shows a fictional two-panel desk stand. It provides teaching material for the workflow below. Use the following facts to see how the brief turns into step cards:
Task: Assemble the example stand and illustrate phone placement.
Parts: One back panel (A) and one base panel (B).
Starting point: Separate panels, with both insertion slots clear.
Orientation: Both white front markers face the reader.
Sequence: Orient, align, join, position, then place the phone.
Cable: Optional; route it through the lower opening before placement.
Join check: The panels meet without a visible gap.
Unknown: No approved phone compatibility or load limit is provided.Use those facts with the storyboard prompt below and the pictures in the example kit. The table here shows how to map the supplied facts to pictures and result checks. It is an editorial example you can compare with your own draft.
The example's complete five-step sequence
| Action | Picture needed | Result to inspect |
|---|---|---|
| Orient the two panels. | Show the upright back panel, flat base, and both front markers. | The markers face the reader. |
| Align the slots. | Show the back panel above the base with the relevant slots unobstructed. | The slots share the insertion line. |
| Join the panels. | Show the downward movement and its completed state. | The joint has no visible gap. |
| Position the stand and route an optional cable. | Show the flat base and lower cable opening. | The base rests flat; any cable stays clear of the joint. |
| Show the phone's resting position. | Show it against the back panel, centered over the base. | The phone is shown centered, with any cable clear of the joint. |
Build the alignment card from the second row: use the alignment drawing, write the action beside it, and add the result to inspect. The schematic below puts both slots on the same vertical line.
2. Align the panels
Hold panel A above panel B and line up their insertion slots.
Check that the slots share the same vertical line before joining the panels.
If they are misaligned, lift A away and line it up again. Avoid pressing the panels together at an angle.
Continue with the remaining rows, then arrange the cards across pages. Open the example PDF to compare the grouping: inventory and preparation first, assembly next, placement and removal after that, then checks and problem solving. Retain only the sections your own task needs.
The phone image shows the fictional example's intended position. It does not establish compatibility or load capacity for a real stand. Replace the example's facts, drawings, and recovery instructions with information for your own product.
Choose the task and the reading format
Write the completion condition before opening a design tool: “After using this guide, the reader can [perform the task] and confirm [an observable result].” If you cannot finish that sentence, the guide's scope is still too broad. “Learn the app” gives you no clear stopping point. Completing one named task gives you a sequence you can illustrate and check.
Record the reader's starting state as well. Do they have all the parts? Are they already signed in? Do they need an administrator to grant access? Instructions that begin after a hidden prerequisite leave a new user stuck before step one.
Match the format to where the guide will be used
| Reader's situation | Useful format | What determines the space needed |
|---|---|---|
| A short assembly with few parts | A printed card or one-page PDF | Part identification, orientation, and the action must remain legible at the printed size. |
| Setup includes preparation and several visible changes | A short booklet or multipage PDF | Keep related steps together and put the relevant warning beside its action. |
| Someone follows software instructions on a phone | A web page with stacked step cards | Each screenshot needs enough surrounding interface to locate the control without constant zooming. |
| A team consults the instructions beside a workstation | A posted sheet with a separate reference document | Show the actions needed at that station. Put uncommon exceptions in the linked reference. |
| Several product variants have different instructions | A clearly identified guide for each distinct sequence | The reader must know which model or version the pictures describe before starting. |
A four-page layout can work well for a small assembly: preparation, assembly, first use, and problem solving. Those are possible page jobs, not required page counts. If all the necessary instructions fit comfortably on one page, use one page. If the assembly needs a detailed close-up, give that image room.
Keep the full manual available when the quick start covers only part of a product's use. Include its location in plain text as well as a link or QR code. Preserve any required information in the quick start itself; a link does not repair an omitted prerequisite or warning.
Prepare the facts and reference images
For your first guide, collect the task's starting point, the items or access it requires, the action sequence, a picture for each important change, and the result that means the task is complete. If you are documenting something you do yourself, perform it while making this record. For customer instructions, check the sequence with the product or process owner.
Gather the current instructions and product information. Resolve disagreements before asking AI to draft. A photograph can establish what a part looks like; it cannot establish a maximum load, a cleaning chemical's suitability, or a certification.
For software, record the version or capture date, account role, and relevant device. A screenshot from an administrator's account may contain controls that the intended reader cannot use. For a physical product, identify the model and included parts, then follow the manufacturer's sequence and photograph the changes.
Use your own pictures or confirm that you may distribute supplied images in the guide. Record the image owner and permission with the working files.
The fuller brief below is useful when several people or product variants are involved. For a small guide, a source name and page number may be enough to locate an instruction. Use identifiers such as F01 when they make checking easier; keep them in your working notes.
Guide title:
Reader and starting state:
One task this guide covers:
Completion condition:
Product model / software version / account role:
Delivery format and intended reading size:
Sources:
F01 | Fact or instruction | Source, version, and location
F02 | Fact or instruction | Source, version, and location
Prerequisites and included items:
Approved action order:
Visible result after each action:
Known mistakes and approved recovery actions:
Required warnings and where they apply:
Approved care, removal, or reset instructions, if relevant:
Unsupported claims and unresolved questions:
Images:
I01 | File | What it shows | Product/version | Source or owner
I02 | File | What it shows | Product/version | Source or owner
Person responsible for checking the finished guide:Write MISSING_FACT beside a required detail you cannot establish. If the action depends on it, ask the person responsible for the instructions before including that action. Optional information can stay out of the guide. This avoids filling the page with caveats that do not help someone complete the task.
Separate observed problems from predicted problems. A support record or a walkthrough can justify a specific troubleshooting entry. AI can suggest questions to investigate, but its suggestions are not evidence that a problem occurs or that a proposed fix works.
For products involving electricity, heat, chemicals, structural loads, medical use, or other consequential risks, work from the approved technical instructions and the review process applicable to that product. Use AI to organize established information. Have the responsible specialist check technical actions and required warnings.
Ask AI for a storyboard you can inspect
Use a text assistant to organize the sequence and edit the instructions. Choose one that accepts images if you want it to comment on reference pictures. If image input is unavailable, describe what each picture shows and compare the pictures yourself. A guide built from photographs or screenshots does not need an image generator.
A storyboard connects each instruction to the picture that explains it. Ask for a table first. You can spot a missing state or an unexplained transition before spending time on layout. Attach the brief and identify each reference image by filename. Do not assume the model has inspected an attachment merely because it repeats the filename.
Create a visual quick-start storyboard using the brief and reference
images below. Use only the supplied facts for task instructions.
The reader is: [reader and starting state]
The task ends when: [observable completion condition]
The guide will be read: [format, device or paper size]
First list conflicts and missing facts that prevent the task from
being explained accurately. Do not invent the missing instructions.
Then return a table with these columns:
Step number | Starting state | Instruction | Image needed |
Visible result | Approved recovery, if supplied | Source IDs
Use as many steps as the task needs. Keep each step focused on an
action the reader can perform and check. Preserve exact part names,
interface labels, quantities, and orientation cues.
Specify what each image must show, including the view, moving part
or control, relevant detail, and end state. Keep labels, arrows,
numbers, and instructions as separate editable elements.
Separate optional actions and conditional branches from the main
sequence. Use MISSING_FACT where an essential result or recovery
cannot be established. Do not invent a fix to fill a table cell.
Suggest a page grouping that fits the reading format. Explain any
step that needs a second image or a close-up. Do not force a page
count, step count, or number of troubleshooting entries.
Brief and image references:
[paste the brief and attach the approved images]Check the table in sequence. The result of one step must match the starting state of the next. If a part appears assembled before the guide explains how it was joined, add the missing transition. If a software screen changes after a confirmation button, include that action and the resulting screen.
Review each source ID against the cited record. Models can attach a real identifier to a sentence that the source never supports. Read the cited passage and compare it with both the instruction and the requested picture.
If the storyboard becomes crowded, ask the model to separate the main route from optional routes while keeping the conditions for taking each route. Review what moved. Preparation needed by every reader belongs before the main route; an optional accessory belongs beside the step where the choice matters.
Write instructions that a picture can support
Start a step with the action and name the object. In the stand example, “Assemble the panels correctly” leaves the reader to work out which panel moves and what alignment means. “Hold panel A above panel B and line up their insertion slots” gives them an action to match to the picture. The next sentence, “Check that the slots share the same vertical line,” tells them what to inspect before continuing.
When an AI draft is vague, give it a specific revision request: “Rewrite the alignment step using the supplied names A and B. Say which panel is above the other and what the reader should compare. Leave joining for the next step.” Check the revision against the drawing. Asking for a more professional tone will not supply a missing direction or separate two actions.
Microsoft's guidance on procedural writing recommends numbered steps, imperative verbs, and identifying where an action takes place when that could be unclear. It also allows closely related short actions in the same interface location to share a step. Divide instructions at the points where a reader needs to inspect a result or choose a different route.
Put an observable check immediately below the instruction. For assembly, the check might concern the alignment shown in a reference picture. For software, it might be a named screen, saved item, or confirmation state. “Everything is ready” gives the reader no way to compare the actual result with the intended one.
Add recovery only when your source gives a fix for that symptom. Some steps need a stop instruction or a support contact instead of a repair. Do not add a troubleshooting block to every card merely to keep the cards visually identical.
[Step number] [Action heading]
Do: [Specific instruction from the verified source.]
Check: [What the reader can see after completing the action.]
If this check fails: [Approved recovery or escalation, when known.]
Image: [Verified state or action view, with a relevant close-up.]
Source: [Internal reference for the author; omit from customer copy
unless the reader needs it.]Use part names and interface labels consistently. Keep the manufacturer's name or the exact screen label across captions, instructions, and troubleshooting. An image labeled “base” beside text that calls it a “support plate” creates an unnecessary identification problem.
Choose photos, screenshots, diagrams, or AI images
Choose the image that explains the action accurately
| What the reader needs to recognize | Use | Check before placing it |
|---|---|---|
| The actual product, connector, texture, or hand position | A photograph of the correct product state | The relevant feature is visible and the hand does not hide the action. |
| A control, menu, permission, or screen result | A screenshot from the intended software and account role | The label matches the instructions and the crop preserves enough context. |
| Orientation, movement, or a hidden relationship between parts | A diagram based on verified geometry | The simplification retains the feature that distinguishes a correct action from a mistake. |
| A decorative background or a non-instructional concept | An AI-generated illustration, reviewed for its intended use | It cannot be mistaken for a product state or evidence of how an action works. |
For photographs, choose a plain surface and keep the camera position consistent through the sequence. Take an inventory shot, the starting state of each action, and its result. Photograph a detail separately when a wide view cannot show it. Keep the original files so that a later crop can be corrected without recreating the action.
For screenshots, capture the screen at a readable size and crop around the relevant control with enough surrounding interface to locate it. Use demonstration data where possible. Remove private information before sharing the reference with AI, and check that the cleanup has not erased a label or result the reader needs.
For diagrams, duplicate the same part shapes across steps. Change only the position, view, or relationship required by the action. Give parts stable letter labels and a physical landmark, so their identity does not depend on color alone. If you change the camera angle, name the new view and show a landmark that lets the reader orient it.
Arrows should have a defined job. A movement arrow shows a direction of travel; a callout line identifies a feature. Put a short instruction beside a movement arrow when its meaning could be confused with rotation or part identification. Show a before-and-after pair when a static arrow cannot explain the change.
OpenAI's image-generation documentation notes limitations with precise text placement, consistency across generations, and control over structured composition. A generated image therefore needs inspection for the exact task. Keep labels and instructions in the layout editor, where you can correct them without regenerating the picture.
You can ask AI to suggest a crop and callouts before altering a picture:
Inspect the attached reference for [specified step]. Return a text
plan describing the useful crop, the detail that needs a close-up,
and where a separate callout or movement arrow would help.
Use only features visible in the reference or stated in the brief.
Flag any detail that needs another photo. Do not generate or edit
an image. Keep labels and instructions as separate layout elements.Check the suggested crop against the full picture. It must retain the feature the reader uses to identify the part and understand the action. If you later use a generated image, compare it with the reference and replace it if a slot, connector, or part changes.
Inspect image detail at its final displayed or printed size. Start with the original photograph or screenshot; stretching a small preview can leave a connector or button unreadable. For example, a 1,200-pixel-wide picture placed at 4 inches wide has 300 pixels per inch. That calculation does not tell you whether the original picture was focused or whether the relevant feature is large enough to recognize. Check the feature itself in the exported guide.
Build an editable layout without code
The steps below use PowerPoint for Windows as a page-layout tool: one slide becomes one guide page. Start with one action per page so you can check the picture and copy before trying to fit several cards together. You can use the same arrangement in an editor you already know. Keep the picture, instructions, labels, and arrows as separate objects you can change.
Open a blank presentation. On the Home tab, open the New Slide menu and choose a Blank layout. Use this blank slide for your first card and remove any unused title slide. Microsoft's basic PowerPoint instructions cover slides, pictures, and shapes.
Set the page dimensions before adding content: Design > Slide Size > Custom Slide Size. For an exact portrait A4 canvas, enter a width of 21 cm and a height of 29.7 cm. For US Letter, enter 8.5 in by 11 in. Check the dimensions instead of relying on a preset's name. Microsoft's slide-size documentation covers platform differences and custom measurements.
Select Insert > Text Box and draw a heading area near the top. For the example's alignment card, type “2. Align the panels.” Leave space around the edges; 15 mm is a starting choice for a home-printed prototype, subject to the printer's usable area. Add the product or task version and page number at the bottom. Microsoft's text-box instructions show how to place text independently of a template.
Select Insert > Pictures > This Device, choose the picture for this action, and insert it below the heading. Drag a corner handle to resize it. To remove unneeded surroundings, select the picture, use Picture Format > Crop, drag the black crop handles inward, and select Crop again to finish. Follow Microsoft's cropping instructions for the available controls. For the alignment card, both slots and both part labels must remain visible after the crop.
Add separate text boxes below the picture for the action and check. Copy the alignment card's instruction, then its sentence about the shared vertical line. Include the supplied recovery if needed. Select the text and use the Home tab's font controls to set its size. Try 12-point body copy and a larger heading for a first print prototype, then inspect it at actual size. Enlarge small captions instead of squeezing them in.
Add any missing part letters with Insert > Text Box. For a callout line or movement arrow, use Insert > Shapes and draw the chosen shape. Keep it clear of the feature it explains. The alignment card needs a line that identifies alignment; a downward movement arrow belongs to the joining step. Compare every label with the parts list before copying the layout.
In the left thumbnail pane, right-click the finished slide and choose Duplicate Slide. The copy appears after it, as shown in Microsoft's slide-duplication instructions. Change the step number, heading, picture, action, and check. Make sure no caption or recovery from the previous action remains. Check that the new picture starts where the previous step ended.
Make a one-card export using the PDF instructions below. Print it at actual size or open it on the intended device. Can you read the text and identify the important detail without enlarging the whole page? If not, give the picture more space, add a close-up, or enlarge the text. Do this before repeating the layout throughout the guide.
Once the card is readable, arrange the full sequence. On print pages, place two or more cards together only where their pictures and text still fit. For a narrow screen, keep a vertical stack. Put prerequisites before actions, optional routes beside the choice they affect, and completion checks after the last required action. Keep each picture with its instruction. Move a long troubleshooting table to another page if it interrupts the sequence.
Save the editable PowerPoint file, then export the complete guide. Keep the original images alongside it. Give the PDF a task or model name and version, such as
desk-stand-quick-start-v1.pdf, so a recipient can distinguish it from an older copy. Inspect the exported pages again after combining cards; a readable single-card prototype does not guarantee that the final grouping works.
Adapt the guide to the situation
Software setup and first-use instructions
Begin with access requirements and the account role used in your screenshots. Capture the actual starting screen, then the state after each action. Use the interface's exact labels in the instructions. If the application has different desktop and mobile routes, show which route the guide covers before the first screenshot.
A tight crop can make a button readable while making it impossible to find. Pair a wider locating image with a close-up when necessary, and use the same callout number in both. Keep an expected result as text beside the screenshot so the reader can check the task even if minor spacing or styling changes.
Explain what to do when permissions or account status change the route. “Ask your administrator for access” belongs before an unavailable action, with the request details supplied by the process owner. Do not invent a workaround for a control that the reader lacks permission to use.
Several product models or nearly identical parts
Put the model identifier where the reader will see it before opening the sequence. Make a working table that maps each variant to its part inventory, relevant pictures, action order, and completion checks. This reveals whether the variants can share a guide.
Use a shared guide when the same instructions and images remain accurate. Create a distinct branch or guide when the task changes. Avoid instructions such as “use the appropriate part” unless you show how to identify that part. For look-alike components, photograph the distinguishing edge, opening, label, or other identifying feature at a readable size.
Keep variant-specific text and images together in the source file. Replacing only a product name can leave an old connector or orientation in the diagram. During review, compare the inventory and first action for each supported variant.
Optional accessories and conditional steps
Place the choice before the action it affects. In the stand example, tell readers to route a cable through the lower opening before placing the phone; readers without a cable can go straight to phone placement. State the condition, then the route to follow. For an optional accessory, identify the accessory and the step that a reader without it should continue to. Use explicit destinations when a branch spans pages.
Walk through every branch separately. Check that each route reaches the intended finished result and that skipped actions are genuinely optional. If branches keep sending the reader backward or across several pages, split the routes into separate cards or guides. That is a layout decision you can assess from the sequence, without imposing an arbitrary maximum number of steps.
Multiple languages
Finish checking the instructions before translating. Give the translator a glossary of part names and interface labels plus the images for context. Ask AI to preserve quantities, units, negation, and references to numbered parts, then have a fluent reviewer check the meaning against the source.
Lay out each language at a readable size. Translated text may need a different amount of space, so reserve flexible text areas and inspect line breaks in the exported file. Separate language editions can be easier to follow than several tiny translations beneath each picture.
Check direction-sensitive visuals in editions that read from right to left. Reading direction may change the layout; the physical movement shown by an assembly arrow still has to match the product. Check each arrow's direction after arranging the translated layout.
A phone as the main reading device
Try the actual guide on the device someone will use beside the product or application. An A4 page reduced to a phone's width can make a good printed guide difficult to follow. For a web version, stack the step number, image, instruction, and check in a single column.
Keep full-resolution details available when the reader needs to inspect a small feature. Make the main sequence readable without opening a separate image at every step. Check links and branch destinations on the phone, and keep a route back to the main sequence. If the same guide is also printed, prepare a print layout from the same checked instructions.
Black-and-white printing or small package inserts
Print a grayscale prototype. A color difference may disappear, so identify parts with letters and visible landmarks. Inspect arrows against their background and check whether a light border still distinguishes a close-up from the main picture.
For a small insert, test the actual folded size and reading order. Put the first action on the panel the reader is meant to open first. If the full sequence fits only after making labels unreadable, use a larger foldout or booklet. A QR code can lead to additional detail, but the printed instructions must still contain the information required for the task they claim to explain.
If a commercial printer will produce the insert, get its specifications before laying out the final file. Confirm the finished size, fold positions, bleed, and the area where text must stay clear of a cut or fold. Use the printer's PDF requirements and review a proof. A home-printed copy can check reading size, but it cannot confirm the supplier's trimming or folding.
Readers who need more help identifying the starting state
Show the unassembled inventory or the app's initial screen before the first action. Name the objects in both image labels and text. Explain a required tool or access condition at the point where it matters. Do not assume that a beginner will identify a component from a technical noun alone.
During a walkthrough, observe whether the reader chooses the correct object before acting. If they choose the wrong one, improve the identification image or instruction. Adding general reassurance does not fix a missing landmark. If an experienced reader can skip preparation, give that route a clear entry point without removing the preparation a new reader needs.
Removal, reset, and routine care
Include these instructions when they are necessary for the guide's promised task or the instructions you are working from require them. Obtain the actual sequence. Disassembly and software reset may have prerequisites or effects that cannot be established by reversing the setup pictures.
For software, check whether reset affects saved work or settings and state the confirmed effects before the action. For a product, keep the manufacturer's care limits. If the instructions do not establish a cleaner, maintenance interval, or storage condition, ask the owner. AI should not produce a care section simply because the template has an empty space.
Repair common problems in the draft
Symptoms you can inspect while editing
| Problem | What to inspect | Practical repair |
|---|---|---|
| A part changes shape between steps. | Compare it with the original reference and the previous state. | Reuse the original photo or master diagram. Replace the altered state. |
| The reader cannot tell whether to push, lift, or turn. | Look for an ambiguous arrow or a missing view. | Name the movement in the instruction and show a second state or close-up. |
| The diagram contradicts the text. | Check orientation and whether the label points to the named part. | Resolve the instruction against its source, then correct both elements. |
| An instruction uses a control that is absent. | Compare software version, device, and account role. | Capture the intended reader's route or document the required access. |
| A page fits only with tiny captions. | View or print at the intended size. | Remove repeated copy, enlarge the picture, or move related cards to another page. |
| AI adds a confident repair or product claim. | Look up the exact supporting fact, including its scope. | Delete the claim or get it confirmed by the responsible owner before drafting it. |
| The last action has no clear end state. | Compare the storyboard with the completion condition. | Describe the confirmed result and show what the reader should inspect. |
| The PDF differs from the editable file. | Inspect fonts, crops, overlapping objects, and page breaks. | Repair the source and export again. Check the replacement file. |
AI can help identify candidates for these problems. Ask it to compare the draft with the source brief and report the exact sentence, image, or transition that needs attention. Recheck its findings yourself. A review that says the guide is clear without naming what was compared gives you little information to act on.
Export and check the delivered file
Export the pages as pages, not a presentation handout with several slides on a sheet. In PowerPoint for Windows, File > Export leads to Create PDF/XPS Document. On macOS, use File > Save As or File > Export and choose PDF. Microsoft's PDF export instructions explain the platform options. For Windows print output, Standard is the higher-quality choice; inspect the result before choosing a smaller-file option.
Open the PDF in a viewer outside the editor and inspect every page. Look for clipped labels, missing images, changed fonts, displaced arrows, and a card split across pages. Select and copy text to check that instructions have not become one large image. Test all support links and QR destinations using the delivered file.
Accessibility needs a separate review of how the content is structured and used. Selectable text helps, but a PDF also needs appropriate tags and a logical reading order. Follow Microsoft's accessibility export guidance for your version, and inspect the PDF afterward. On supported Mac versions, the electronic-distribution accessibility option uses Microsoft's online conversion service; use an approved conversion route for confidential material.
Adobe's PDF accessibility guidance covers title, language, tagging, and reading-order checks, including items that need manual review. Check that a reader hears a step's instruction and its explanation in a useful sequence. An automated pass cannot establish whether a diagram explains the intended action.
For a web guide, provide the essential instructions as real text alongside the images. A short alternative text can identify a diagram; complex relationships need a fuller text equivalent. W3C's guidance on complex images explains how to supply the information that the image conveys. If the adjacent instruction already explains the movement and result, write the image alternative to complement that text.
For digital content, WCAG's text-contrast criterion sets a 4.5:1 minimum for normal text and 3:1 for qualifying large text, with stated exceptions. Its non-text contrast criterion covers meaningful graphical information at 3:1 against adjacent colors, subject to its scope and exceptions. Check relevant arrows and outlines as well as copy. These contrast checks alone do not establish accessibility.
Have someone follow the guide without coaching
Use the intended product or application, the final exported guide, and the reader's actual starting conditions. For printed instructions, print at the intended size. For a phone guide, use the phone. Reviewing the source file on a large monitor does not tell you whether the delivered version is readable in use.
Ask someone who represents the intended reader to attempt the task. Explain the task's goal, then let the guide provide the instructions. Record where they hesitate, choose the wrong part, miss a branch, or cannot recognize completion. If you must intervene, note the exact point and what information was missing.
Use a review record with a row for each observation:
Guide version and delivered format:
Product / app version / account role:
Starting conditions:
Step or branch:
What the reader did:
What they could not identify or understand:
Relevant source and proposed change:
Change checked by:
Result of the next walkthrough:Fix the cause of the observation. A wrong orientation needs a better landmark or view. A skipped confirmation needs an explicit action. A missed optional route needs a clearer branch location. “Read carefully” shifts the burden to the reader without improving the instruction.
Ask the product or process owner to check factual accuracy and any required technical review as well. A successful reader walkthrough does not establish safety or regulatory compliance. Keep the reader's completion result separate from the owner's approval of the information.
Use the example kit and keep your guide current
The Pinefold review kit contains the example's facts, diagrams, PDF, and optional build scripts. You can open the example and inspect its pictures without running the scripts. The saved test results report checks of that example's files; they do not show a customer completing the task.
The kit's product brief, QA checklist, and original prompt describe its four-page example. Use the adaptable brief and prompt in this article for your own sequence.
Keep the source brief, reference images, editable layout, exported guide, and review record together. When a part, screen label, permission requirement, or instruction changes, identify the affected steps and replace their text and images. Export again, check those steps and their connected branches, and make the current guide available at the location your readers already use.
