agentleFS
brandonbryant12brandonbryant12/clearframe/skills/clearframe-review/SKILL.md

clearframe-review

Skill1 starsChanged 23 days agoMITView on GitHub

Turn a person's feedback on a ClearFrame cut into bounded, previewed edits — record notes against the revision they watched, apply the edit, return a before/after candidate, and record their acceptance or rejection. Use for rough-cut reviews, "at 2:13 this is confusing", "keep the voice", "cut this sentence", restoring an earlier version, or any edit to a film someone has already watched.

The file

We split this skill into the rules it gives an agent

3 rules. About 0 are common, 0 are distinctive, and 3 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈0

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 3

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Write the paper edit: paper DIR (add --suggest-cuts only if the person wants cuts proposed).

    line 12
  2. Very specific:

    Read the output: the beat, the quoted words, any "near the cut to …" warning.

    line 29
  3. Very specific:

    Ask only for material ambiguity, a keep conflict (override DIR --keep k002 --by NAME --said "…" records their permission), missing …

    line 46

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Review and edit a ClearFrame film

The contract is docs/editing.md; the conversation rules are in docs/collaboration.md. Everything lives in the project's review/ folder. Commands are pnpm cf ….

Before the first look

  1. Write the paper edit: paper DIR (add --suggest-cuts only if the person wants cuts proposed). For a recording, the words stay as recorded unless they ask.
  2. Make the rough cut with what exists: blocks, basic motion, captions, and "placeholder": "what will go here" on beats whose picture isn't made (canvas/art elements can be "unfinished": "note"). Don't draw bespoke pictures for every beat and don't run reviewer rounds yet.
  3. draft DIR --rough renders it, saves revision r00N and writes review/index.html. Send the page; ask what feels wrong, where, and what to keep.

--rough relaxes only declared stand-ins (see docs/editing.md): a clipped source line, label or caption is still an error. Fix those; never mark finished text unfinished to get past the audit.

When notes arrive

For each note in the reply:

pnpm cf note DIR "the diagram is confusing" --at 2:13 --rev r003 --by NAME [--keep voice]
pnpm cf note DIR "[ClearFrame r003 @ 2:13.40 · s047 · “…”] the diagram is confusing"   # pasted from the page
pnpm cf notes DIR --import notes-r003.json                                            # downloaded from the page
  • --rev is the cut they watched (without it, the newest cut with a video is assumed: say so if unsure).
  • Read the output: the beat, the quoted words, any "near the cut to …" warning. If the person may have meant the neighbouring beat and it matters, ask; otherwise add --beat with your reading.
  • "Keep the voice/audio/words/this picture/the look" becomes --keep on the note (scoped to its beats) or keep DIR … --by NAME --said "…" for wider scope. Read the semantics in docs/editing.md before cutting anything under a voice or words keep.
  • Then notes DIR: anything stale or orphaned goes back to the person with its old revision, time and words. Never apply a note to whatever is at that timestamp now.

Making the edit

  1. Edit only the note's beats (notes DIR shows where each note is now). Picture notes: edit storyboard.json. Recording cuts: never edit vo or a beat's WAV (a voice keep compares every slice, sample for sample, with the master); use cut:
    pnpm cf cut DIR --note n014 --by NAME                     # the sentence the note points at
    pnpm cf cut DIR --words "which is a story for another day" --by NAME
    pnpm cf cut DIR --pauses-over 1.2 --keep-pause 0.5 --by NAME
    pnpm cf cut DIR --paper review/paper-edit.md --by NAME    # words they struck with ~~…~~
    --by NAME means the person asked for the cut; --agent means you decided it (a voice keep refuses agent cuts). --dry-run shows what would go. --note and --at --rev find the exact words they pointed at by identity, not by text: if those words were already cut or the note is stale, nothing is cut and you should ask the person again.
  2. revise DIR --note n014 checks the edit stayed inside the note's beats and broke no keep, saves a candidate revision, renders before/after passages and writes review/compare/A-B/index.html. If it reports changes outside the scope, revert them or widen with --scope ID,ID --reason "…" (say why to the person).
  3. Reply in one or two lines: what you changed, what else moved or looks different (from the report), and the compare page. Generated narration: a changed line re-records its whole take; say so and check the budget before voice without --draft.

Ask only for material ambiguity, a keep conflict (override DIR --keep k002 --by NAME --said "…" records their permission), missing authority, or paid work beyond the budget. Otherwise apply and show.

Recording their answer

pnpm cf accept DIR r006 --note n014 --by NAME --said "yes, much clearer"
pnpm cf reject DIR r006 --note n014 --by NAME --said "no, the old one read better"
pnpm cf restore DIR r003 --by NAME --said "go back to Tuesday's cut"
pnpm cf decide DIR r006 --checkpoint rough --reason "one-shot: all notes applied"
  • accept/reject/restore --by are only for what the person actually said; quote them. Silence, elapsed time or "looks fine I guess" about something else is not acceptance.
  • Keeps on a recording hold through splits and merges: a refused cut has written nothing; ask the person rather than looking for another path.
  • reject undoes the candidate where nothing changed since (beat order included) and reports conflicts; restore brings back a whole revision (setting aside later library overrides and shadowing media under review/aside/, then verifying the result) or --beats (refused when it would change media other beats use, unless --shared, or replay merged recording), saving the current state first. uncut DIR --cut c004 undoes one recording cut exactly.
  • For the final, record acceptance of the whole cut (accept DIR rNNN --checkpoint final --by NAME --said "…"): accepting one note's result does not approve the film, and any later change reopens it.
  • In one-shot work use decide; it is shown as an agent decision, never as acceptance.

Reporting

revisions DIR, diff DIR rA [rB], checkpoints DIR (Rough cut and Final close on decisions; coverage by chapter for long films), runlog DIR (measured time only; gaps between commands are not work). When you summarize, separate: rendered (a revision exists), applied (a candidate exists), accepted (a person said so).

The studio keeps feedback available on Final exports. Use Whole-cut note for overall pacing, tone or direction; use + Note for a pinned moment. Never attach overall feedback to timestamp zero. Reopen threads when the reviewer requests changes; final export quality does not mean the person accepted the cut. Review the selected version in the Storyboard grid.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.