← All craftbooks

Edit Notes Pass

Media: audio and videoevalv1.0.0released 2026-06-05workflow: build-loop

Steps

Entry step: scope. Each step names the specialist role it wants; the full working prompt is expandable.

1. Scope the reviewregisseurentry

   footage inventory, transcript read, review criteria locked

   Show working prompt

   Define the edit pass before forming a single opinion. Step 1: read media/index.json from the workspace — every clip with its id and duration — and read the transcript under transcripts/ plus the storyboard or brief if one exists, so you know what the piece is trying to be. Step 2: lock the review criteria — every keep-or-cut call carries a timecode and a stated reason, pacing judgments name the specific stretch that drags or rushes, b-roll gaps are named where the footage cannot carry the point, and no note references footage that is not in the index. Step 3: define what the cut list must record for each decision — which clip, the in and out points, the action (keep, cut, trim, reorder), and a short note tying it back to the edit notes. Step 4: write an acceptance-criteria checklist ('every note carries a timecode', 'every cut decision has a reason', 'every cut-list row points at a real clip', 'in and out points sit inside that clip's duration', 'the draggy stretches are addressed, not just observed'). Call write_task_note with the clip count + criteria + checklist and write notes/scope.md. No reviewing yet.

2. Write the edit notesreviewer

   timestamped pacing and cut notes, clip by clip

   Show working prompt

   Review the footage against the locked criteria. Step 1: walk the transcript and the media index clip by clip, in story order, and write timestamped notes — what works, what drags, where the energy dips, where a cut or trim would help. Every observation carries a timecode in mm:ss form and names its clip. Step 2: for each stretch that drags, say exactly what to do about it — trim from where to where, cut entirely, or reorder — and why the piece is better for it. Step 3: name the b-roll gaps: the moments where the words need cover the footage does not have, so the owner can plan pickups. Step 4: keep opinions honest and concrete — 'the intro runs long' is a note, 'trim 0:12-0:38, the point lands at 0:39' is an edit. Write the notes to notes/edit-notes.md and call write_task_note with the count of cut decisions and the biggest pacing problem found.

3. Build the cut listdeveloper

   the machine-readable record of every edit decision

   Show working prompt

   Distill the edit notes into the cut list the assembly step can execute. Step 1: for every decision in notes/edit-notes.md, write one record to cutlist.json as a JSON array — each record with the clip id exactly as it appears in media/index.json, the in and out points, the action (keep, cut, trim, or reorder), and a short note carrying the reason over from the edit notes. Step 2: verify precision as you go — every clip id must exist in the media index, and every in and out point must sit within that clip's duration; a cut list that references invented footage or impossible timecodes is worse than no cut list. Step 3: order the records in final-piece order so the assembly reads top to bottom. Step 4: re-read the result once as JSON — valid, complete, no decision from the notes left behind. Call write_task_note with the record count and the total runtime the kept material adds up to.

4. Evaluatereviewer

   Grade the deliverable against every acceptance criterion. All pass → finish; any fail → loop back and fix the gap.

   Show working prompt

   Open cutlist.json, notes/edit-notes.md, and media/index.json and verify every criterion from notes/scope.md: every cut-list record names a clip id that exists in the media index, every in and out point sits inside that clip's duration, every record's action traces back to a timestamped note with a reason, the notes address the pacing problems rather than just observing them, and the b-roll gaps are named. Write PASS/FAIL per criterion; on any failure, name the offending record or note and loop back to cutlist.
   
   Then route — this is the whole point of the loop:
   
   - **Every criterion PASSES →** call `advance_task_step({ ref, stepId: "evaluate", next: "finish" })`.
   - **Any criterion FAILS →** write the specific gaps to notes, then call `advance_task_step({ ref, stepId: "evaluate", next: "cutlist" })` to loop back. The builder fixes exactly those gaps.
   
   Never route to `finish` while any criterion is unmet. The build phase's completion gate already blocked a grossly-incomplete deliverable; your job is the judgment an automated check cannot make (does it actually work, read well, look right). After ~3 unproductive loops, stop and report DONE_WITH_CONCERNS so the user can step in.

5. Finishdeveloper

   All acceptance criteria met. Stamp a short summary and report DONE.

   Show working prompt

   Every acceptance criterion passed. Write a one-paragraph DONE summary to task notes via `write_task_note`: what was built, the deliverable path(s), and a one-line confirmation that each criterion is met. Then report DONE.

Triggers

Phrases that suggest this craftbook to a crew.

• edit pass
• review the footage
• make a cutlist
• what should we cut
• tighten the video

Source

View this craftbook on GitHub · MIT license