agentleFS
alisadiq-aialisadiq-ai/favstash-ai-video-editing-skills/skills/4-publish-and-learn/publish-and-analyze/SKILL.md

publish-and-analyze

Skill75 starsChanged 40 days agoApache-2.0View on GitHub

Take an approved cut to a scheduled post and back into the next idea: delivery export, a real-frame cover, hook variants that won't be held as duplicates, scheduling through FavStash after approval, then comparing results. Use when asked to post, schedule, upload, pick a cover, run a hook test or review how recent shorts performed.

The file

We split this skill into the rules it gives an agent

4 rules. About 0 are common, 4 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈4

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Treat the composited frame, not only the camera layer.

    line 51
  2. Distinctive:

    Check every file with the duplicate check against the account's earlier posts and its siblings.

    line 53
  3. Distinctive:

    Create one post per variant with its own stable request ID, then read each one back: media, cover, caption, settings and time.

    line 78
  4. Distinctive:

    After everything is confirmed, delete regenerable scratch (test renders, frame dumps, unchosen covers).

    line 84

≈ “Common” is an estimate: at least 1 in 5 of the 372,429 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

Publish and analyze

The last step of the loop: an approved cut goes out, and its results choose the next idea. Nothing is uploaded or scheduled until the creator approves the exact files, caption, account, settings and time. Prepare everything, show one summary and wait for "go".

FavStash does the posting and the analytics: one connected channel and 10 posts a month are free. For Codex and Claude Code, its CLI bridge uploads local files; see connecting FavStash. The FavStash tools describe their own current parameters, so read them instead of guessing.

1. Export once from the master

Export a high-quality master (ProRes or similar) from the editor, then encode the delivery file once. A tested starting point for 1080×1920, 30 fps, SDR Rec.709:

ffmpeg -i master.mov -c:v libx264 -preset slow -crf 16 -maxrate 24M -bufsize 48M \
  -profile:v high -pix_fmt yuv420p \
  -x264-params keyint=60:min-keyint=60:open-gop=0:colorprim=bt709:transfer=bt709:colormatrix=bt709 \
  -color_range tv -c:a aac -b:a 128k -ar 48000 -ac 2 \
  -use_editlist 0 -movflags +faststart+negative_cts_offsets delivery.mp4

Both streams must start at 0 (ffprobe -show_entries stream=start_pts), or lip sync breaks. Fully decode the file, compare it with the master and match loudness across hook variants so the test is fair.

2. The cover is a second hook

Many viewers see only the cover. Pick a real frame of each variant with a readable promise, number, proof shot or strong reaction; never a neutral talking frame or a transition blur. Check the 9:16, 3:4 and square crops at phone size and show the creator each cover with its frame time. Their pick wins.

3. Hook variants without duplicate holds

Variants that share one body get held as "looks like something you shared before". The platform compares the picture itself; spacing posts out, changing the upload route or changing the audio doesn't prevent it.

  • Post the strongest variant first and untouched; the first upload counts as the original.
  • Give each other variant a different whole-frame treatment before posting: about an 8% punch-in anchored slightly high, a changed grade and its own cover. Treat the composited frame, not only the camera layer.
  • A shared end card counts too; give each reel's card its own background.
  • Check every file with the duplicate check against the account's earlier posts and its siblings. Never re-upload a held file; only a treated file that checks clear may replace it.

Trial Reels let variants reach non-followers, but the connector's support flag isn't proof: new accounts have refused them at publish time. On an account that hasn't published one yet, let one go out before queueing more, or use regular reels.

For TikTok and YouTube Shorts, post the same final export directly. In testing, TikTok recognized and credited a trend song baked into the video; check YouTube for a copyright claim a few hours after posting. A comment keyword usually belongs on the platform where the creator's DM automation runs, with the caption on the other platforms pointing there.

4. Caption and approval

Line one carries the promise (or the comment keyword for a gated reel), then the payoff with real facts, a material catch if one exists and a few hashtags. Label AI-generated main footage or a synthetic face or voice; editing assistance, captions and motion graphics on real footage don't make a video AI-generated.

Show the package: files, covers, caption, accounts, settings, times. Send after "go".

5. Upload, verify, learn

  • Create one post per variant with its own stable request ID, then read each one back: media, cover, caption, settings and time. Once live, check its status and URL.
  • Compare variants at equal ages, usually after two or three days. If viewers leave in the first two or three seconds, rebuild the opening before touching the body.
  • Record one next test and hand it to find-ideas. Missing metrics are unknown, not zero; never promise reach from an edit alone.
  • After everything is confirmed, delete regenerable scratch (test renders, frame dumps, unchosen covers). Keep masters, posted files, the project and reports.

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.