We only use cookies to enhance your experience. No cross-site tracking, no ads. Read our Cookie Policy.
Explain how it should work, once. Your agent writes the stories your developers need to build it.
Show how it should behave, where things live, and the thinking behind it.
Paste the share link. Your agent reads the transcript and the frames.
They land in your tracker, scoped and labeled, ready to pick up.
Record a video explaining how the feature should work, the flow through it, and the decisions behind it. Your agent reads the recording, pulls the design from Figma, and the stories land in your tracker.
Connected to your agent.
Connected to your agent.
The Figma design on screen while you record.
Connected to your agent.
Claude, Cursor, or similar.
Hit record and walk a developer through the feature as if they were sitting next to you. Supercut transcribes your narration as you go, so anything you say out loud becomes a field the agent can fill in. Cover the six things a developer will otherwise come back and ask you:
What someone is trying to do, and the path they take. Show it in motion.
What changes, what triggers the change, and what is not allowed.
Empty, loading, error, long strings, missing data.
What you chose, what you ruled out, and why you landed there.
Which screens are in this handoff, and which are not.
The Figma file. Measurements and tokens come from there, not from the video.
The developer watching sees the motion. Your agent reads your words and still frames, so narrate what moves and what you are pointing at. It will write "the dropdown" unless you say it is a Select.
Paste the Supercut link and the Figma link and prompt your agent to read it all and create dev-ready stories on your tracker. From the Supercut connector, your agent can pull what it needs: the transcript, the exact frames. Access is permission-aware, so it only sees recordings you can see.
Your agent turns the walkthrough into separate stories. It reads the flow and the reasoning from your recording, pulls the exact values from Figma, and creates each story through your tracker. For each story it writes the title and description, adds a frame from the moment you explained it, and links the recording (where developers can ask clarifying questions if they want to).
Record your walkthrough covering the entire design handoff, then paste this to your agent. Swap [SUPERCUT LINK] for your recording and [TEAM OR PERSON] for whoever is building it.
Here is a Supercut recording of a design walkthrough: [SUPERCUT LINK]
Read it and create dev-ready stories in Linear for [TEAM OR PERSON].
1. Fetch the transcript for that recording. It comes back as sentences with start and end timestamps in seconds.
2. Work out how many separate pieces of work the handoff covers, and create one story per piece of work rather than one story for the whole video.
3. For each story, pull a frame from the second where I am explaining that piece. Use the transcript timestamps to pick the moment.
4. Write each story with these fields:
- Title: what needs building
- Scope: what is in this story, and what is explicitly out
- How it works: the flow, and the path someone takes through it
- Rules: what changes, what triggers the change, what is not allowed
- Edge cases: empty, loading, error, long strings, missing data
- Decisions: what I chose, what I ruled out, and why
- Reference: a link back to the recording, and the frame you pulled
5. Assign it to [TEAM OR PERSON].
Rules:
- Use only what I actually said or showed. If I did not cover a field, write "not specified in the handoff" instead of filling it in yourself.
- For spacing, type, color and token values, pull them from Figma using the Figma MCP. Do not infer them from the frames.
- The frames are stills, not video. Do not describe hover, focus or motion unless I said it out loud.
- Do not invent component names.
- Quote me directly in the Decisions field, so nobody reopens a settled decision later.
- When you are done, list the stories you created and anything in the recording you could not turn into one.The agent can only fill fields you gave it. Cover the six elements above while you record and every field lands automatically.
A written spec cannot show a flow in motion. A live walkthrough can, but only for the people who joined the call. An async video shows the prototype working, and everyone on the product team watches it in their own timezone. Cover the whole handoff in one pass and nothing is left to guesswork: design and engineering work from the same picture. Then the video is the start of the task, not the end of it. Your agent reads it, pulls the design from Figma, and creates the tickets in seconds.
Flodesk's design team records three or four walkthroughs before lunch, explaining how a feature should work across the US, Europe and Vietnam. Read how they do it.
Show the new component once, file the work everywhere it lands.
This is one of several things your agent can do across the two tools.
Record the bug happening. Your agent gets the transcript and frames, and the Linear ticket writes itself.
Point your agent to your PM's feature demo video, and it writes the help center article, screenshots included.
Record yourself doing the work once. Your agent picks it up and writes the documentation directly in your database.
Record your update. Your agent posts the summary, the action items, and links to the moments that matter.