agentleFS
lorena-bordonaba-paulorena-bordonaba-pau/youtube-agent/.claude/skills/channel-kickoff/SKILL.md

channel-kickoff

Skill2 starsChanged 49 days agoMITView on GitHub

Build the channel's strategy from zero, or evaluate the strategy of an existing channel — business and objective, audience profile, adjacent interests, provisional content pillars and differentiation. Guided for someone who knows nothing about YouTube strategy. Triggered by "/kickoff", "start the channel", "I'm new, where do I begin", "build my strategy", "evaluate my channel's strategy". In Spanish, "empezar el canal", "soy nuevo, por dónde empiezo", "hazme la estrategia", "evalúa la estrategia de mi canal".

The file

We split this skill into the rules it gives an agent

6 rules. About 0 are common, 4 are distinctive, and 2 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.

  • 2

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Propose, then confirm: When you can infer a sensible answer, offer it as a proposal ("From what you said, I would put the audience at 25-45.

    line 37
  2. Distinctive:

    Never write it down unconfirmed.

    line 39
  3. Very specific:

    No OAuth token → own analytics are unavailable.

    line 53
  4. Distinctive:

    Ask what the channel sells: a course or community, a service, a product, or nothing (audience, sponsors, authority).

    line 57
  5. Very specific:

    Put what the channel does next to what the strategy says it should do.

    line 121
  6. Distinctive:

    Every cell in "what there is" carries its source and date.

    line 133

≈ “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 kickoff: the strategy

WHEN

The first thing anyone does with this harness, and whenever the creator wants the strategy rebuilt. tools/init.py points here while memory/strategy.md says NOT POPULATED.

It covers strategy only: what the channel is for and who it speaks to. Competitors come after, in /competitors, validated channel by channel with the creator. Ideas, titles, thumbnails and scripts are flows the creator starts on their own. Do not drift into them here, however tempting.

THE METHOD

Strategy runs in one direction: what you sell → who buys it → what that person wants → what you will talk about. Every later decision — which competitor to watch, which idea to record — is checked against these four answers. A channel with views from people who will never buy has a strategy problem, not a growth problem.

HOW TO TALK TO THE CREATOR

They may know nothing about strategy. So:

  • One question at a time, in plain words, never a form.
  • Offer options with an example each, plus "something else". Nobody answers "what is your awareness level?"; everybody answers "does your viewer already know they have this problem?".
  • Say in one line why you are asking. People answer better when they know what the answer changes.
  • Propose, then confirm. When you can infer a sensible answer, offer it as a proposal ("From what you said, I would put the audience at 25-45. Right?"). Never write it down unconfirmed.
  • "I don't know" is a valid answer. Record it as unknown; do not invent it.

EXECUTION ORDER

0. Pick the mode — measure, do not ask

python3 tools/yt_channel_stats.py
python3 tools/yt_recent_videos.py --limit 50 --stats
  • Fewer than 5 long-form videos, or no channel yet → new channel: the audience profile is a declared hypothesis.
  • 5 or more → existing channel: everything the creator says is checked against data, and the kickoff ends with an evaluation (step 7).
  • No OAuth token → own analytics are unavailable. Say so, work as a new channel, and note that the evaluation needs data/auth/ set up.

1. Business and objective Ask what the channel sells: a course or community, a service, a product, or nothing (audience, sponsors, authority). Then who pays (consumers, small businesses, large companies) and roughly what it costs (tens, hundreds, thousands). Then the single objective: leads, authority, reach for sponsors, or sales. Why it matters: a high-ticket B2B service needs few, qualified viewers; a low-ticket community needs volume.

2. Audience profile — seven fields, one question each: gender, age range, geography (and the vocabulary of that market: words, money, references that make that viewer feel spoken to), awareness level (does not know the problem / knows it / is comparing solutions), goals (what they want to become or achieve), pains (what stops them), and who the channel is NOT for.

In existing mode, measure before asking and show the contrast:

python3 tools/yt_demographics.py --days 90
python3 tools/yt_geography.py --days 90

The gap between the audience the creator believes they have and the one they have is usually the first finding. Record both.

3. Adjacent interests Propose 5-8 topics outside the core niche that the same viewer already cares about (someone into running also cares about nutrition, injuries, gear), and let the creator keep, drop or add. These feed the market map in /competitors as search terms, so write each one the way a viewer would type it — in the audience's language and in English, and in the language of any other market that is ahead in this niche. Other markets are not noise: a topic already winning abroad and absent at home is the best opportunity this map can find.

4. Provisional pillars Four umbrellas, drawn from the core niche plus the strongest adjacent interests, and checked against goals and pains:

  • Distinct from each other — a video fits one pillar, not two.
  • Named by what the viewer walks away with, never by a tool: the tool is a topic inside the pillar and goes stale first.
  • Each one must serve the objective from step 1. A pillar that brings the wrong viewer is dropped, however popular.

They stay provisional: the market has not backed them yet. /competitors will.

In existing mode, hand the catalogue to the youtube-strategist agent: it classifies the last 50 long-form videos into these four pillars and returns views, retention and mismatch per pillar. That measurement goes into step 7.

5. Differentiation One question: "What can you say that others in your niche cannot, or will not?" — experience, results, a stance, a format. If the answer is "nothing", say so plainly: it is the most important gap to close, and /competitors helps find it.

6. Write it down

  • memory/strategy.md: every section, dated, each audience line marked measured or hypothesis, and **Stage:** strategy. Delete the NOT POPULATED line.
  • CLAUDE.md section 1: fill the brackets from steps 1-2 and delete the "FILL THIS IN" quote, if it is still there.
  • config/config.json: channel_context as a consultant's brief (who, what they sell, what sets them apart), and language / region. Show the creator the diff before writing, and write only what they approve.

7. Evaluation — existing channels only Put what the channel does next to what the strategy says it should do. Write it to memory/strategy_eval.md, dated:

DimensionWhat there is (measured)What would be rightGapAction
ObjectiveWho the content attracts todayWho the objective needs
AudienceDemographics and geography from StudioProfile from step 2
Market vocabularyWhat titles and topics sound like todayThe audience's words
PillarsCatalogue split and mismatch per pillarThe four pillars
AwarenessLevel the current topics assumeLevel the audience is at
DifferentiationWhat the channel offers todayStep 5

Every cell in "what there is" carries its source and date. A dimension that could not be measured says so instead of guessing. End with the three gaps that matter most, in order.

SOURCES

Everything the creator says is config-grade: their declaration, not a measurement. Studio demographics and geography are youtube_api. The pillar split from the strategist is derived. The evaluation never mixes them without labelling which is which.

OUTPUT

  1. A short summary of the strategy in five lines: sells, to whom, objective, the four provisional pillars, differentiation.
  2. The files written, and what each one now says.
  3. In existing mode, the evaluation table and the three gaps that matter most.
  4. One next step: "Run /competitors to map the channels already winning with this audience and back the pillars with real videos."

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.