FIX WHAT IS BROKEN

Hand your agent a bug you cannot describe

There is a class of bug that survives every bug report, because the thing that is wrong is the sequence, not any single frame of it. A list flashes the old data before the new. A modal closes itself if you click too fast. A field resets when you go back. Written down, those all sound like nothing. Recorded, the order of events is right there.

Quick answer

To report a bug you cannot describe, record the sequence that breaks at the speed that triggers it, and call it out the moment it happens. Paste the Clipy link into your AI agent, such as Claude Code, Codex or Cursor. It reads the order of events from the timestamped transcript and key moments, and tells you where the behavior diverges from what you expected.

Product walkthrough. Tap for sound.

What to record

  1. Get into the exact state that precedes it, including anything that seems irrelevant.

  2. Reproduce it at the speed that actually triggers it, fast or slow.

  3. If it is intermittent, keep going until it happens, and say so when it does.

  4. Say what the correct order of events would have been.

Say this while you record

Call it the moment it happens. "There, that flash" gives the agent a timestamp to anchor on, which is much more useful than describing the flash afterwards.

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 a recording of an intermittent bug. The problem is the sequence, not any single screen. Walk me through the order of events as you understand them from the recording, with the timestamps, and tell me at which point the behavior diverges from what I said I expected.

Then look for the code that could produce that ordering: race conditions, stale state, effects firing twice, optimistic updates that are not reconciled. Tell me your top two candidates and how you would tell them apart.

Asking for candidates rather than a fix is deliberate. For a timing bug the first plausible explanation is often the wrong one.

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 downRecording it instead
A bug that is about orderingThree paragraphs that still do not convey the order.The order is the recording.
An intermittent failure"It happens sometimes", and nobody can reproduce it.One captured instance, with everything that led up to it.
The agent's first guessBased on your theory of the bug.Based on what happened, which is often not your theory.

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 the failure, not the attempt

    If it took nine tries, that is fine, keep the recording running. But say clearly when the real one happens, because otherwise the agent has to work out which of the nine attempts was the interesting one.

  • Timing bugs need the real speed

    Slowing down for the camera is the one thing that will make the bug not appear. Do it at the speed that breaks it, and narrate afterwards if you have to.

  • Paste the link once the transcript shows

    The transcript, summary and key moments appear on the watch page moments after Stop.

Common questions

What if it did not happen while I was recording?

Then you have a recording of it working, which is genuinely useful too: it tells the agent what the passing path looks like. Keep it and record again when it fails.

Can the agent tell how much time passed between two events?

Yes, within the resolution of the transcript and the key moments. The transcript carries timestamps and the key moments are anchored to instants in the recording, so relative timing is available. Do not treat it as millisecond instrumentation.

Is this better than a screen recording pasted into a chat?

A plain video file gives an agent very little it can read. The point of an agent-readable recording is that the same capture also produces text and frames the agent can actually work with.

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.