REVIEW, VERIFY AND HAND OFF
Turn a walkthrough into a written spec
Writing a spec is slow because it is a translation job: you already know what you want, and the work is converting it into sentences. Recording yourself explaining it over the real product skips the translation, and the agent does the writing. You go from explaining to correcting, which is much faster than explaining to writing.
Quick answer
To turn a walkthrough into a written spec, record yourself explaining the feature over the existing product, covering the unhappy paths and what is out of scope. Paste the Clipy link into an AI agent like Claude Code, Codex or Cursor. It writes the spec with the states and unhappy paths you mentioned, marks anything ambiguous as an open question, and writes no code yet.
What to record
Open the part of the product the feature touches.
Explain what you want while pointing at the things it affects.
Cover the unhappy paths out loud: what happens when it fails, when it is empty, when there are too many.
Say what is explicitly out of scope.
Say this while you record
Out of scope is the most valuable thing you can say, and the thing people leave out. Say it.
The prompt to paste
Replace the link with your own recording's share URL, then paste this into Claude Code, Cursor, Codex, or whichever agent you connected.
Read https://clipy.online/video/<id> It is me explaining a feature over the existing product. Write it up as a spec: what it does, the states it has, the unhappy paths I mentioned, and what I said is out of scope. Mark anything I left ambiguous as an open question rather than deciding it yourself. Do not write any code yet. When the spec is agreed, we will use it as the brief.
"Mark ambiguity as an open question" is the whole value here. A spec that quietly guesses is worse than no spec.
What the agent gets from that link
A summary of what happened
A short written account of the recording, so the agent starts with the shape of the request rather than a wall of raw text.
A timestamped transcript
Your narration with times against it, when audio was captured and processed. This is where your reasoning lives, and it is usually the part that decides whether the agent fixes the right thing.
Key moments with frames
Selected instants from the recording with an image of the screen at that point, so the agent can look at the state you were describing instead of guessing at it.
Pointer and click evidence
Where you clicked and what you were pointing at, with the control's accessibility label when the capture recorded one. It turns "this button" into a specific element.
What changes
| Writing it down | Recording it instead | |
|---|---|---|
| Getting the spec written | An afternoon you keep postponing. | Five minutes of talking and a draft to correct. |
| Unhappy paths | Remembered during implementation, badly. | Prompted for while you are still looking at the screen. |
| Ambiguity | Resolved silently by whoever builds it. | Listed as open questions before anything is built. |
| Handing it to someone else | They read the spec and miss the reasoning. | They read the spec and can watch the recording behind it. |
Tips
Connect your agent once
Run the setup at /agent-setup once. After that every Clipy share link works in the same session you are already in.
Record over the real product
Explaining a feature in the abstract produces an abstract spec. Pointing at the screens it touches anchors every requirement to something that exists.
Do not let it start building
The prompt says not to write code. Keep it that way until you have read the spec, because correcting a document is much cheaper than correcting a branch.
Open questions are the deliverable
If the spec comes back with no open questions, either you were unusually clear or the agent guessed. Check which.
Common questions
Is a recording a good format for a spec on its own?
No, and that is the point. The recording is the input. The written spec is the artifact, and it is reviewable in a way a video is not.
How long should the walkthrough be?
A few minutes for most features. If it is running past ten, the feature probably needs splitting, which is a useful thing to learn early.
Can I use this for a bug report backlog instead?
Yes. Record one pass over several problems and ask for a written list. It is the same job with a different output shape.
Other jobs
The same job, elsewhere on Clipy
This page is the job. These cover the same ground for a specific agent, a specific team, or as a tool you can open right now.
The full list is on the use-case index. If your agent is not connected yet, start at agent setup.