Reap source

Skill jperrello/C0BALT_CUT/.claude/skills/reap-source

Investigation on using Claude Code to automatically generate profitable YouTube videos.

Install

npx -y skills add jperrello/C0BALT_CUT --skill reap-source

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

2 things to look at

no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.
2 stars2 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Reclaim disk by deleting a PROCESSED source's heavy artifacts — the full podcast source.mp4 (re-downloadable from its saved URL), every clip_NN.* stage intermediate, the b-roll cutaway cache, and the source's mcptube ingest (via broll-cleanup). KEEPS all lightweight JSON (ingest/transcript/topics/segments/grade/broll_plan) as the on-disk memory and updates the sources-ledger registry. Manual only — never auto-runs in the pipeline. Refuses a source with no finished shorts in output/ unless --force. The "cleanup" half of the disk-hygiene pair (sources-ledger is the "memory" half). Deterministic, non-fatal, idempotent (re-reaping a reaped source is a no-op).

SKILL.md

2.9 KB, 616 tokens by cl100k_base, as published. Nobody here has run it

reap-source

work/ accumulates ~25× the size of the finished shorts: the full podcast download and every per-stage clip_NN.*.mp4 linger forever, while the deliverable is already safe in output/. This reaper deletes the heavy reclaimable stuff and keeps the cheap JSON that records what was done.

Usage

reap-source.sh <id>             # reap one source (e.g. 81d2d55a40 or work/81d2d55a40)
reap-source.sh <id> --dry-run   # show what would be freed, delete nothing
reap-source.sh <id> --force     # reap even with no finished shorts in output/
reap-source.sh --backlog        # reap every source whose shorts are in output/
reap-source.sh --backlog -n     # dry-run the whole sweep (recommended first)

What it deletes vs keeps

Deletes (heavy, reclaimable):

source.mp4 — the full podcast; re-downloadable any time from the URL in ingest.json
clip_*.{mp4,mov,wav,m4a,webm,mkv} — every per-stage intermediate (the bulk of the bloat)
broll/ — the cutaway cache, plus the source's mcptube ingest (db/frames/chroma) via broll-cleanup

Keeps (cheap, the memory):

all *.json — ingest.json, transcript.json (expensive whisper output), topics.json, segments.json, every *.broll_plan.json, *.grade.json
all *.txt / *.path / *meta sidecars

After reaping it writes a .reaped marker (date + bytes freed), calls sources-ledger record <id> to flip the registry entry to status:reaped, then a final sources-ledger sync.

Safety

Manual only. Nothing in the pipeline auto-reaps. The decision is yours.
Gated. Refuses a source that has no finished output/<slug>/*.mp4 unless --force — so a failed/half-run source isn't reaped before it ever delivered.
Re-downloadable. To restore a reaped source: bash start.sh work/<id> (or re-ingest from its saved URL). transcript.json is kept, so re-transcription is skipped on a re-run.
Idempotent. Re-reaping an already-reaped source is a no-op. Non-fatal — per-file errors don't abort a --backlog sweep.

Restore a reaped source

bash start.sh work/<id>     # re-ingests source.mp4 from the saved URL; transcript.json is reused