agentleFS
lorena-bordonaba-paulorena-bordonaba-pau/youtube-agent/.claude/skills/competitor-ideation/SKILL.md

competitor-ideation

Skill2 starsChanged 48 days agoMITView on GitHub

Next-video ideas drawn from real outliers among competitors and inspirations. Triggered by "ideas for this week", "what should I record", "what's blowing up in my niche", "analyse my competition". In Spanish, "ideas para la semana", "qué grabo ahora", "qué está petando en mi nicho", "analiza a mi competencia".

The file

We split this skill into the rules it gives an agent

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

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈3

    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. Very specific:

    If memory/strategy.md is NOT POPULATED, say that ideas without a strategy are guesses and offer /kickoff first.

    line 13
  2. Distinctive:

    Idea first, outlier second: Start from something the creator genuinely wants to say, then find a format YouTube has already proven and fit …

    line 31
  3. Distinctive:

    Same title, different video: Only for titles so general that nobody owns them ("if I had to start X today, this is what I would do"), and …

    line 34
  4. Very specific:

    Read memory/strategy.md, memory/creator_profile.md and memory/MEMORY.md.

    line 77
  5. Distinctive:

    The order adapts: If step 1 returns a 10x outlier, prioritise it and go deep even though the manual said transcribe everything first.

    line 88

≈ “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

Ideation from competitors

WHEN

Next-video ideas, competitor analysis, "what is working in the niche", the weekly opportunity review. Always started by the creator.

If memory/strategy.md is NOT POPULATED, say that ideas without a strategy are guesses and offer /kickoff first. If it is at **Stage:** strategy, the pillars are not market-backed yet: offer /competitors, or go ahead and label the ideas as resting on provisional pillars.

HOW IDEAS ARE BUILT

Every idea belongs to a pillar. Plan in batches of 8 — two per pillar — so that after a month each pillar has been tested twice. Then review with /pillars: learn, iterate, and scale what worked. One idea per pillar is too little to conclude anything.

Three ways to get from an outlier to an idea. Name which one each idea uses:

  • Bring it over. A title structure or format proven in another niche or another market, applied to this channel's topic. The structure travels; the topic is yours. The strongest source is a winner abroad with no equivalent in the creator's language.
  • Idea first, outlier second. Start from something the creator genuinely wants to say, then find a format YouTube has already proven and fit the idea into it. The idea is original; the packaging is not a gamble.
  • Same title, different video. Only for titles so general that nobody owns them ("if I had to start X today, this is what I would do"), and only when the creator brings something the original could not. Otherwise it is a copy and it competes with the original.

EXECUTION ORDER

  1. Outliers across the configured channels python3 tools/yt_outliers_channels.py --min-ratio 2.0 --sample 30 Reads the competitor and inspiration lists from config/channels_lists.json. If very few come back, drop to --min-ratio 1.5 before concluding there is nothing.

    If there is a competitor board (memory/competitor_board.md holds its URL), sync it with the competitor-board skill and use its data.top_organic_outliers too. It covers every long-form video of the last 6 months, not a sample, and it has already dropped the videos whose views look bought as ads.

  2. Hand-saved outliers python3 tools/yt_outliers_playlist.py --days 7 Whatever was saved during the week. If the playlist is not configured the tool says so with exit code 4; that is not a failure.
  3. Filter for relevance before spending quota Discard by title anything that does not fit the channel. Only then: python3 tools/yt_transcript.py --video ID --plain on the ones that do. Transcribing everything is wasted time.
  4. Validate each idea python3 tools/kw_research.py --kw "idea 1; idea 2; idea 3" heuristic score: it ranks the ideas against each other, it does not measure absolute demand.

4.5 Size each idea python3 tools/topic_tam.py --topic "idea 1::regex" --topic "idea 2::regex" Match the audience's interest, not the exact case (an agent that runs your YouTube channel is sized as "an AI that runs your business"). It gives the TAM the topic moved on the board and floor / expected / ceiling views for this channel. heuristic: quote it as an estimate, with the user's own record on the topic next to it.

  1. Prioritise against the channel, not in the abstract Read memory/strategy.md, memory/creator_profile.md and memory/MEMORY.md. An idea that attracts a viewer the objective does not need gets dropped, however strong its outlier. An idea with a good outlier that does not fit the channel's archetype gets dropped, and you say why.

Before building on any outlier, check its likes per 1k views against the channel's usual. Views that arrive without likes are paid reach (ads), and the ratio they post says nothing about the format or the title. Affiliate links are no signal: almost every channel carries them.

The order adapts. If step 1 returns a 10x outlier, prioritise it and go deep even though the manual said transcribe everything first.

SOURCES

Steps 1-2 derived (the ratio is computed over each channel's mean; the board uses the median). The board's ad_suspect flag is heuristic. Step 3 youtube_api. Step 4 heuristic — declare it.

OUTPUT

Prioritised ideas, grouped by pillar. Each with:

  • The pillar it tests and the technique it uses.
  • The outlier backing it: title, channel, ratio and URL verbatim from the tool.
  • Why it worked there and what changes when bringing it to this channel.
  • Your own angle, not a copy.

At the end, a single recommendation and the reason for it.

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.