A simple Sudoku workflow for people

Let your AI chat see the next move.

Share the Sudoverse Sudoku Agent Skill, upload a photo or screenshot of your current Sudoku board, and ask the AI to explain one verified human-style step at a time.

See the 3-step setup About the API

Start with one message

Watch the workflow happen.

Paste the skill link into an AI chat, attach your Sudoku image, and ask for one verified human-style step. This small walkthrough shows what the conversation looks like.

The preview link opens the exact board explanation in a browser when the chat cannot render the visual directly.

AI chat · exampleReady

Please use the Sudoverse Sudoku Agent Skill:
https://playfulbit.com/sudoku-agent-skill/SKILL.md

Read the current board from the image, use the Sudoverse Agent API to find one verified human-style next step. Show me a visual explanation of the technique, and do not guess. If you create a preview, include the “Open the visual explanation” link.

I’m reading the board, checking the transcription, and requesting one verified action from the Sudoverse Agent API.

Here is the visual explanation.

The preview shows the human technique, full candidate context, and the exact placement or candidate removal.

Open the visual explanation

From puzzle image to helpful hint

You stay in control of the board. The AI reads the image, checks uncertain cells, and asks Sudoverse for a verified next move.

Share the skill

Give your AI chat the public skill link so it knows how to use the Sudoku Agent API.

Upload your board

Attach a clear Sudoku image. A photo, screenshot, or scan works well.

Ask for one step

The AI transcribes the grid, checks ambiguity, and requests the next human-style action.

Follow the highlight

See a placement or candidate elimination, then update your board and continue.

53··7····6··195····98····6·8···6···34··853··17···2···6·6·····28···419···5····8··79

example result

Not just an answer

Your AI can say: “This is a Naked Single. Place 5 in R5C5 because it is the only candidate left.” For advanced patterns, it can highlight the linked cells and show which candidate to remove.

What you can ask for

learn

“Give me the next step, not the whole answer.”

The default workflow returns one verified placement or candidate elimination so you can learn while solving.

visualize

“Show me where to look.”

The agent can use the API’s target cells, highlights, and links to create a lightweight static board or short technique animation.

preview

“Open the visual explanation.”

The agent can return a shareable preview link with the board, full candidates, technique context, and highlighted placement or elimination. Open an example preview.

continue

“I applied that move. What is next?”

Upload an updated image or provide the updated grid and ask for another step.

trace

“Teach me the next five steps.”

The agent can request a bounded trace when you want a guided sequence instead of a single hint.

Tips for a better image

Use a straight, well-lit image

Keep the full 9×9 board visible, avoid glare, and crop out unrelated page content when possible.

Let the AI ask about unclear cells

A blurry digit or pencil mark can change the puzzle. The workflow should show suspected transcriptions instead of silently guessing.

What if the chat cannot render the board?

Ask the agent to return the “Open the visual explanation” link. The hosted preview works in a browser and the API can also produce a temporary HTML preview retained for 24 hours.

Use the official skill link

Share the public skill: open SKILL.md. The source package is also available on GitHub. The AI still needs vision support to read the uploaded image.