agentleFS
indranilbanerjeeindranilbanerjee/cutforge/skills/revision-round/SKILL.md

revision-round

Skill0 starsChanged 27 days agoMITView on GitHub

Turn client notes into a revision round: map, decode, fix minimally, report with timestamps. \"here are my notes on the video\"

The file

We split this skill into the rules it gives an agent

3 rules. About 0 are common, 1 is distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈1

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 2

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Fix the class, not just the instance: When one name tag was wrong, re-check every name tag.

    line 35
  2. Very specific:

    Never overwrite the delivered version: a new VER per round.

    line 47
  3. Very specific:

    If the notes are for a project that has moved to the client's machine, work in that copy (config.json folders), and follow the same order.

    line 51

≈ “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

Revision round: every note mapped, fixed, re-checked, and reported with timestamps

Client notes arrive as voice notes, chat lines, timestamps, screenshots or one vague sentence ("abrupt stops in many places"). Treat each as a precise bug report: find where it is, what mechanism causes it, and what the smallest correct fix is. Never re-render the whole film for a list of local fixes.

Workflow

  1. Capture every note in NOTES_vN.md (template in the project) exactly as given, one row each: | # | At | Note (client's words) | Shot | Class | Fix | Done |. Split compound notes. Keep the client's words; your interpretation goes in Fix.
  2. Locate: python tools/qc/where.py 8:26 12:03 ... maps film times to chapter, shot, the line being spoken and the media on screen. Times from the client may refer to a preview of an older version: check which file they watched.
  3. Classify each note: fact (identity, name, date, relation, ritual), story (order, missing beat, tone, balance), picture (layout, timing, text), sound (abrupt, level, music), voice (pronunciation, delivery), technical (playback, file). Facts first: they are the most damaging and the cheapest to fix.
  4. Decode vague notes into mechanisms (references/feedback-decoder.md): "abrupt" → tails, fades, bridges, fixed shots; "not in the right place" → chronology table, late insertions; "primitive" → engine, sound sources, motion design; "slideshow" → framing and transitions. Run the matching QC tool to find every instance, not only the one the client noticed.
  5. Ask only what cannot be resolved: batch the remaining questions (one round) with stills of the shots in question. Do not ask what the data answers.
  6. Plan the smallest rebuild: list the chapters touched, which narration lines regenerate, which cues change. Order: facts.md/decisions.md → script.json → gen_vo.py <ids> → film/timeline.py → shots.json / custom chapters → stills of changed shots → FRESH=1 pipeline.sh <changed chapters> → film_mix.py <changed chapters> (plus the chapter where any long cue starts; chapters that moved are re-mixed automatically) → VER=v(N+1) assemble.sh → QC.
  7. Fix the class, not just the instance. When one name tag was wrong, re-check every name tag; when one chapter end had dead air, run qc_audio over the whole film.
  8. Verify with the video-qc layers for the changed chapters plus the full-film checks (verify --final, qc_audio, qc_story). Mark each NOTES row Done only after its fix is seen in a still or heard in the mix.
  9. Report back in short plain sentences: what changed, with new timestamps ("8:26 the family page now names Grandpa"), what was checked and how, anything not done and why, and the one question still open. Send the new preview and, when useful, the storyboard page for the changed shots.

Rules

  • A late insertion (a new photo, a new beat) re-runs the story-order table and the story editor re-reads the whole chapter, not only the new shot.
  • Never overwrite the delivered version: a new VER per round; keep NOTES_vN.md for every round.
  • A note that conflicts with an earlier decision is raised with the client, quoting the decision, not silently applied.
  • Record every changed fact with its source in facts.md, every changed choice in decisions.md.
  • If the notes are for a project that has moved to the client's machine, work in that copy (config.json folders), and follow the same order.

References

  • references/feedback-decoder.md: what clients say → the mechanism → where to look → the fix
  • ${CLAUDE_PLUGIN_ROOT}/docs/lessons-learned.md sections C7, E2, I5

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.