“Why didn’t this video work?” asks the assistant to guess too much. “Where do we explain the trip twice before showing the packed bag?” gives us something to check. We’ll use that packing-video example to go from an analysis request to a short memo for an editor.
Kabo, an AI agent for creator research, analyzes a supported public video. Our job in Muse is to turn broad edit advice into a claim we can locate and inspect.
Connect Kabo before asking for fresh evidence
Use the Muse connection guide to connect https://kabo.sh/mcp-pub and sign in. Ask Muse to show the ready Kabo tools before requesting fresh research.
Prepare the video link and the decision to review
[ ] Public video URL — The exact YouTube, Instagram or TikTok video to inspect. [ ] Editing question — Repetition, promise, explanation order or another bounded concern. [ ] Original material — Our narration and footage notes, pasted or attached as supported. [ ] Payoff — What the video is meant to demonstrate. [ ] Analysis state — An existing analysis ID, or readiness to review a new-analysis proposal.
Make the question smaller than “why did this video fail?”
A travel creator has a public packing video that feels repetitive. They want to understand whether the explanation circles back before the viewer sees the packed bag.
Tell Muse what you want to understand and which material is yours. Ask about repetition, proof or explanation order. Public analysis cannot reveal the platform’s distribution decisions or private audience retention, and asking a narrower question makes a useful answer more likely.
Here is my public video: [URL]. My audience is [audience], and the payoff is [result]. I want to understand [specific editing question]. Propose the Kabo analysis and explain what evidence it should return. Wait for my confirmation before starting a new saved analysis.
Get the completed evidence, then leave tool state behind
Ask Muse to propose the public URL and analysis question. A new saved Kabo analysis requires Pro and confirmation. Once it is started, the result needs to be completed and readable before we use it. If an existing analysis answers the question, read that first.
The MCP analysis guide explains identifiers, retries and terminal states. Here we’ll focus on what to ask when the completed answer gives us a vague editing judgment.
Challenge “slow” with three narrower questions
Suppose Muse says our packing video feels slow. Ask: which passage repeats information, what is the repeated information, and what would be lost if we remove it?
Those questions distinguish an unnecessary setup line from context the viewer needs. If the result has no usable time range, the honest output is an unlocated observation to inspect manually—not a precise cut invented for the memo.
You described the packing section as slow. Identify the specific repeated information, the supporting passage or returned time range, and whether this is spoken or visual evidence. Explain what the viewer would lose if we removed it. If the evidence cannot locate the repetition, mark the suggestion for manual inspection instead of inventing a timestamp.
What an actionable answer needs to add
Compare these authored examples:
- Too vague: “Cut the slow introduction.”
- Useful shape: “The trip setup is stated again before the packing demonstration. Check the supported passage, keep the baggage constraint, and test removing only the repeated trip description.”
- Still missing: the exact cut range, if the analysis did not return usable timing.
The second answer identifies a review task. It still needs the original footage before it becomes an edit.
Ask for a memo built from the decisions we accepted
Once we agree on which observation is supported, ask Muse for a short memo. Each item needs the proposed change, evidence, context to preserve and inspection still required. Do not let a new list of general pacing advice bury the one change we actually reviewed.
Use only the editing observations we accepted. For each, give the proposed change, supporting passage/time range if available, context to keep and any manual inspection or new narration needed. Separate unconfirmed suggestions from approved edit notes. Do not claim a video file was changed or that retention will improve.
That timestamp was not in the returned evidence. Remove the exact range, retain the question about repeated setup and tell the editor what to inspect in the original. Leave the supported observations unchanged.
Review the accepted edit memo beside the video
Review the memo beside our own footage in Descript or the editor we already use. Confirm the cut in context before applying it. If the video needs a different concept rather than a local edit, continue with the Muse ideas workflow.
Worked examples are illustrative. Product screenshots show existing Kabo web results, not a run of the example in this guide. Use your own materials and the evidence returned in your task.