- Home
- Skills
- Technical Documentation
- Screenshot Guide Studio
More screenshots
Works with the AI Agents you already use
Screenshot Guide Studio
Turn approved screenshots into clear, editable English guides with precise highlights, useful checks and web, PDF and Markdown outputs. Free.
Free
Screenshot Guide Studio
Example session with this skill installed
Use the five approved, already-sanitized Northstar training screenshots and recorded interactions from the software-report example to create a complete English guide.
Audience: workspace analysts. Goal: create a September revenue report, export its CSV and verify the retrieved file. Choose the Product direction. Preserve exact interface labels, highlight the real controls, include a closer view when useful, and keep observed actions distinct from missing evidence. The account contact is already masked; use only the sanitized copies. Generate the web guide, editable JSON, Markdown, annotated PNGs and PDF where the available environment supports them. Label this as a fictional local training example.
- Read your context and instructions
- Compiled the screenshot guide studio
- Generated the data export
Creator-verified local example
The following is the complete written guide from the locally generated reporting example. This listing preview names the sanitized screenshot sources; the generated HTML embeds those actual images. These observations concern a fictional training app, not Agensi buyer-side execution or a real financial account.
Your first report.
From workspace to CSV.
Create a September revenue report, check the preview and confirm the exported file. One clear task, with the exact screens you will use.
Audience: Workspace analysts
Environment: Northstar / Local training app
Updated: 2026-10-18
Fictional training example. Observations apply only to the local demo.
Before you begin
- Open the fictional Morrow workspace with the Analyst role.
- Use the Sample ledger already provided in the training app.
- Allow the browser to save a local CSV when you request the export.
1. Open Reports
From Your workspace, choose Reports in the left navigation.
Screenshot source: reports-01 — Open Reports: the actual local training interface before this check or action.
Expected result: The Reports page lists Monthly revenue overview and Account activity, with a Create report button.
If needed: If Reports is absent in your real app, ask its workspace owner to confirm your role. That permission path has not been tested here.
Evidence (observed): Observed in the fictional local training app on 2026-10-18. This does not verify any third-party product or production account.
2. Create a new report
Choose Create report at the top right of the Reports page.
Screenshot source: reports-02 — Create a new report: the actual local training interface before this check or action.
Expected result: The Create a report screen opens with report name, metric and reporting period fields.
If needed: Do not edit an existing report when you intend to create a separate one. Return to Reports if you opened the wrong item.
Evidence (observed): Observed in the fictional local training app on 2026-10-18. This does not verify any third-party product or production account.
3. Review the settings, then run the preview
Confirm Report name is September revenue, Metric is Revenue, and Reporting period is September 2026. These values are prefilled in this demo. Choose Run preview.
Screenshot source: reports-03 — Review the settings, then run the preview: the actual local training interface before this check or action.
Expected result: September revenue opens with Preview ready and the three Sample ledger rows.
If needed: If your intended period or data source differs, stop and correct it before interpreting the preview. No missing-data error path was exercised in this demo.
Evidence (observed): Observed in the fictional local training app on 2026-10-18. This does not verify any third-party product or production account.
4. Check the numbers and export
Check Direct 12,000, Partner 8,000 and Referral 5,000 USD. Their sum is 25,000 USD. Confirm September 2026, then choose Export CSV.
Screenshot source: reports-04 — Check the numbers and export: the actual local training interface before this check or action.
Expected result: The app displays CSV download requested and the filename september-revenue.csv.
If needed: A download message is not proof that a file arrived. Continue to the browser downloads check before calling the export complete.
Evidence (observed): Observed in the fictional local training app on 2026-10-18. This does not verify any third-party product or production account.
5. Confirm the downloaded file
Open your browser downloads, locate september-revenue.csv and inspect its content. Confirm the header Source,Revenue (USD), the three source rows and Total,25000.
Screenshot source: reports-05 — Confirm the downloaded file: the actual local training interface before this check or action.
Expected result: The retrieved training CSV contains Direct 12000, Partner 8000, Referral 5000 and Total 25000.
If needed: If no file arrived, inspect the browser download status before requesting the export again. Do not treat the confirmation screen as a recovered file.
Evidence (observed): The creator retrieved the CSV through the browser download event and read all five lines. The screenshot shows the request screen; file-content verification was a separate local check. Recipient retrieval is untested.
Finish check
The downloaded CSV opens and matches the three fictional ledger rows and the 25,000 USD total. The screen title alone is insufficient. This is a local training result, not a verified live financial workflow.
Actual retrieved training CSV
Source,Revenue (USD)
Direct,12000
Partner,8000
Referral,5000
Total,25000
Generated deliverables and editing
| Output | Purpose |
|---|---|
| index.html | Self-contained guide with embedded sanitized screenshots, full context, detail views and an editor |
| guide.json | Editable declarative guide data |
| guide.md | The complete written guide |
| guide.pdf | Seven-page PDF from the local builder |
| media/reports-01.png through reports-05.png | Five sanitized screenshot sources |
| steps/01-reports-01.png through 05-reports-05.png | Five numbered, annotated step images |
In the browser, edit the guide or a step, select a focus region and apply the draft. Save web guide exports the applied HTML immediately. Save project includes that HTML, JSON, Markdown, sanitized media and rebuild instructions. Rebuild to synchronize PDF and annotated PNG outputs. Reading marks stay personal and are not exported.
Verification boundary
Actual CSV retrieval, creator-local generation, editing and browser downloads were tested. Source screenshot pixels were re-encoded and the declared contact mask was checked. No external recipient task or live financial workflow was tested. Agensi buyer-side supporting-file execution and generated-file retrieval remain unverified; platform review is requested before approval.
september-revenue.csv
CSV · data export
Example file from a real run - the skill writes it into your workspace.
Connects securely to your tools. The creator never sees your data.
About this skill
Turn approved screenshots into a guide someone can actually follow.
Screenshot Guide Studio creates English how-to guides for software onboarding, website handoffs, team training and practical SOPs. Keep the real interface visible, make the next action clear, and explain how the reader can check the result. The complete workflow is free, with no purchase or rating gate.
Clear steps, useful context
- Preserve exact interface labels and the proportions of the supplied screenshots.
- Combine an action, expected result and relevant recovery note for each step.
- Highlight a precise focus area, keep the full screenshot available and enlarge small controls when needed.
- Separate observed actions, owner-provided instructions and missing verification. A screenshot or toast does not establish a later success.
Make the guide yours
Choose Product, Editorial or Field as a visual starting direction. Edit the title, summary, reader, prerequisites, environment and finish check, then refine the individual steps, screenshot descriptions and enlarged details.
Select a focus region directly on the full screenshot: drag in either direction, or use the keyboard to move and resize it. Numeric coordinates remain available. Apply or discard each draft independently, without losing another step's pending edits.
Changing an instruction or expected result marks an earlier observed step as needing verification. Changing the environment does the same for the guide's observed steps. Original observation records remain historical evidence.
Useful generated outputs
- One self-contained HTML guide with embedded sanitized screenshots and an editor.
- Editable guide.json and portable Markdown.
- Annotated PNG step images and an inspected PDF when those formats are generated.
- A browser-exported project ZIP with the current HTML, JSON, Markdown, sanitized media and rebuild instructions.
Save web guide includes applied edits immediately. PDF and annotated PNG files need a rebuild after editing. Personal reading marks stay in the current tab and are excluded from exports; they are not recipient acknowledgement or proof of task completion.
Three complete worked examples
The creator's local package includes software reporting, a website image handoff and a team task handoff: 16 browser-observed steps across three fictional applications. The reporting example includes an actually retrieved training CSV whose rows total 25,000 USD. The cases demonstrate the workflow rather than claiming compatibility with a real product or customer account.
Requirements and data handling
The deterministic builder needs Python 3.10+ and Pillow. PDF additionally needs ReportLab. The Agent needs suitable image-inspection and file/code-execution tools. Browser tools are useful for authorized capture and visual checks. No paid API, secret key, mandatory backend or external font download is required by the builder.
The builder re-encodes approved PNG/JPEG inputs and removes pixels only in explicitly declared opaque redaction regions. It does not automatically discover private information. Inspect screenshot pixels, captions and instructions before sharing. Source originals stay outside recipient output folders.
Validation and platform delivery
Version 0.2.0 passed local editing, actual browser downloads, project rebuild, mobile layout and nine core contract checks. The three native PDFs contain 22 pages; their complete text and bounds were checked, with representative pages visually inspected.
Agensi supplies Skill instructions through its agent connection. This listing does not promise a downloadable Skill package. The HTML/PDF/JSON files described above are intended generated user work, distinct from exporting Skill instructions. Agensi buyer-side supporting-file execution and generated-file retrieval have not yet been verified; platform compatibility review is requested before approval. Direct offline reopening of the HTML and an external target-user pilot also remain unverified.
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Technical Documentation.
No automatic PII discovery or privacy certification: redaction regions must be specified and inspected, and guide text must also be checked. No background recording extension, automatic tracking of application changes or guaranteed recovery from missing screens. English guide output. Only approved PNG/JPEG inputs are supported by the builder (up to 20 MB and 20 million pixels each). PDF/PNG outputs require a rebuild after browser edits. The three worked cases use fictional local apps; no production save, message delivery, recipient acknowledgement or external target-user task completion is established. Direct offline HTML reopening was not browser-tested because file URLs were blocked by the test browser. Agensi supporting-file execution and generated-work retrieval remain unverified; approval is requested only after platform compatibility review.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before buying a skill.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
1 use
Times used to date
Free to use
No account required to browse
Trust & safety
Security scan
Security scan passed 1 month ago
- Free with an Agensi account
- Software
Python 3.10+PillowModern web browser
- Optional
- ReportLabBrowser automation
ReportLab makes the PDF; skip it with --no-pdf. Browser automation is only for authorized screenshot capture.
Permissions
Read only owner-approved screenshot and guide-data files; write newly generated guide outputs to an agreed new folder. Terminal/code execution runs the declarative builder and export checks. Browser use is for authorized capture and guide inspection. Network access is only needed when the user authorizes visiting an online app or installing missing public dependencies; the generated reader has no required external runtime resource. No credentials, connected service account, environment-variable secret or mandatory hosted backend is required. Do not copy or export Agensi Skill instructions; platform supporting-file execution and generated-output retrieval still require verification.