agentleFS
berkayburakkberkayburakk/tubecrew/skills/channel-pulse/SKILL.md

channel-pulse

Skill2 starsChanged 29 days agoMITView on GitHub

Check how a YouTube channel is doing from its own Analytics - recent videos, where views come from, the searches that bring viewers, whether YouTube tested each new video or cut it - and end with one priority diagnosis. Use for "how is my channel doing", a weekly check, before a topic round, or when a decision's T+30 or T+90 row is due.

The file

We split this skill into the rules it gives an agent

7 rules. About 0 are common, 4 are distinctive, and 3 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.

  • 3

    Very specific

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

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

    If there is no OAuth client yet, follow the upload-private skill's references/oauth-setup.md.

    line 21
  2. Distinctive:

    Before a topic round: the search terms and the "tested and cut" pattern say what to fix (packaging, opening) and what not to blame (the …

    line 36
  3. Very specific:

    When a decision in the ledger reaches T+30 or T+90 (see produce-video, references/learning-loop.md).

    line 38
  4. Distinctive:

    CTR is not in the API: When the diagnosis asks for impressions or click-through rate, ask the creator to read them in YouTube Studio > …

    line 53
  5. Distinctive:

    Every number comes from the tool: If a figure you need is not in the report or the JSON, it is unknown.

    line 60
  6. Distinctive:

    Add it to the script (with a test) rather than computing it by hand.

    line 61
  7. Very specific:

    When a T+30 or T+90 row is due, fill it only from a pulse-<date>.json (and retention-by-scene for the curve), name the file and date in the …

    line 68

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 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

Channel pulse

One command reads the channel's own Analytics and ends with one diagnosis, not a list. The script does the counting; your job is to read the diagnosis, say what it rests on, and turn it into the single next action.

$SKILL below is this skill's folder.

Setup

Same OAuth client and the same read-only analytics login as retention-by-scene (one login serves both):

python $SKILL/scripts/pulse.py --login --profile mychannel

If there is no OAuth client yet, follow the upload-private skill's references/oauth-setup.md.

Run

python $SKILL/scripts/pulse.py --profile mychannel                    # last 10 videos, last 28 days
python $SKILL/scripts/pulse.py --profile mychannel --videos 5 --days 90 --out notes/

It prints the report and writes pulse-<YYYY-MM-DD>.json to --out (default: the current folder). Keep those files: they are the source for the decisions ledger. --json prints the same JSON.

When

  • The creator asks how the channel is doing, or a weekly check.
  • Before a topic round: the search terms and the "tested and cut" pattern say what to fix (packaging, opening) and what not to blame (the topic).
  • When a decision in the ledger reaches T+30 or T+90 (see produce-video, references/learning-loop.md).

Read it in this order

  1. Too early? Videos under 72 hours or under 100 engaged views are not judged. If nothing can be judged, the diagnosis says so and gives the date to come back. Say that and stop; do not interpret noise.
  2. The diagnosis. It is the first real problem in a fixed order: never tested, tested and cut, low discovery, weak opening, no subscriber growth, healthy. Report it with its "based on" line (how many videos, which window). references/diagnosis.md explains each step and its thresholds.
  3. Per-video patterns. "Tested and cut" (a burst of a few days, then a quarter of the peak or less) means YouTube found people for the topic and viewers did not click or stay: fix packaging or the first 30 seconds, not the topic. "Never tested" means the feeds never picked it up: packaging or topic. "Still growing" and "tapering off after a run" need nothing.
  4. Search terms. These are searches that actually brought viewers. Title and tag ideas start here, before any keyword tool. An empty list does not mean nobody searched: YouTube hides rare terms.
  5. CTR is not in the API. When the diagnosis asks for impressions or click-through rate, ask the creator to read them in YouTube Studio > Analytics > Reach. Do not guess them.
  6. Audience. Age, gender and device print only when YouTube returns enough groups; "too little data to interpret" means exactly that. One returned age group showing 100% is the privacy threshold, not the audience.

Rules

  • Every number comes from the tool. If a figure you need is not in the report or the JSON, it is unknown. Add it to the script (with a test) rather than computing it by hand.
  • Analytics lags 2-3 days. Data API view counts are current; a gap between the two is lag, not a drop.
  • One video is one observation. The diagnosis says so when it rests on a single video.
  • Findings are warnings, not gates. The creator may knowingly ignore one; the script never exits 1.

Fill the decisions ledger

When a T+30 or T+90 row is due, fill it only from a pulse-<date>.json (and retention-by-scene for the curve), name the file and date in the row, and write the platform with every view count. A row whose numbers are not in a tool output stays _open_. Never edit a filled row; a correction is a new row.

Report back

One sentence on where the channel stands, the diagnosis with its single next action, what it could not see (too-early videos, CTR, hidden search terms), and the ledger rows you filled. Not a list of twenty numbers.

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.