agentleFS
jsathjsath/claude-skills/skills/social/idea-bank/SKILL.md

idea-bank

Skill14 starsChanged 35 days agoMITView on GitHub

Replaces a creator's ideas meeting: takes five creators in the user's niche, finds their videos that beat their own normal views by 3x or more, explains the topic, hook or format behind each, rewrites the best in the user's own experience with the original named, drops any idea the user cannot speak to from real experience, and ends with a month of video ideas in filming order, each with a hook. Use when a creator wants a month of content ideas, a calendar built from proven videos, or asks \"what should I film this month\". Public data and pasted numbers only; never logs in to any account.

  • Reads credentials

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:

    Before acting as an agent, read its file in agents/.

    line 51
  2. Distinctive:

    Every view count, date and baseline names its source and capture date.

    line 62
  3. Distinctive:

    Never write "I did X" for anything not in EXPERIENCE.

    line 64
  4. Very specific:

    Never ask for passwords or log in to any account.

    line 65
  5. Distinctive:

    ASK questions only the user can answer.

    line 82

≈ “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

Idea Bank

Five agents: scout outliers in your niche, explain why they worked, make your version, drop what you cannot back with real experience, and plan the month in filming order.

Unlike a one-off viral breakdown, this keeps an EXPERIENCE list of what you have actually done, filters every idea against it, and always ends in a month CALENDAR.

What you need

DataClaude CodeClaude.ai
YouTube views and datesYouTube Data API v3 with your key in .env as YOUTUBE_API_KEY (public data only; setup in agents/01-scout.md)Pasted list or screenshots of the channel's Videos tab
Instagram / TikTok viewsPasted numbers or profile-grid screenshotsSame
Transcriptsyt-dlp --write-auto-subs --skip-download <url> for public YouTube videos (the Data API cannot download other channels' captions)"Show transcript" on the video page, pasted
Filming days (optional)Google Calendar connector, read onlySame, or typed

Missing numbers become a question or NOT ENOUGH DATA, never an estimate.

Setup (start, or when PROFILE is empty)

Three rounds, accept "skip", show PROFILE and EXPERIENCE back, then suggest full run.

Round 1, niche

  1. Your niche and viewer, in one sentence.
  2. Platforms and formats (shorts, reels, TikTok, long-form).
  3. Five creators you watch in your niche, with platform for each.

Round 2, what you have actually done 4. Things you have built, tested, sold, fixed or failed at in this niche. Rough notes are fine. 5. Which numbers can go on camera? Which stay private? 6. Topics you never cover.

Round 3, the month 7. Videos per week, filming days. 8. Fixed dates this month (launch, event, travel). 9. YouTube API key in Claude Code, or pasting numbers?

Commands and routing

CommandRuns
full run1 > 2 > 3 > 4 > 5, one summary
scout / patterns / adapt / filter / plan month1 / 2 / 3 / 4 / 5
add experience: <what you did>appends a dated line to EXPERIENCE, re-runs 4 on ASK and DROPPED ideas
swap <date>replaces one calendar slot with the top backup

Before acting as an agent, read its file in agents/. In Claude Code you can copy them to .claude/agents/ to run as real subagents.

  • agents/01-scout.md: baseline = median of last 20 to 30 same-format videos; outlier = 3x+
  • agents/02-pattern-finder.md: topic, hook or format driver; cross-creator patterns
  • agents/03-make-it-yours.md: your angle on each pattern, original named
  • agents/04-honesty-filter.md: KEEP / ASK / DROPPED against EXPERIENCE
  • agents/05-month-planner.md: slots, ordering rules, batched filming days

Safety rules (every agent)

  • Outliers are judged against each creator's own baseline, never raw views across creators.
  • Every view count, date and baseline names its source and capture date.
  • Adapt, never copy: no reused footage, script lines, graphics, thumbnails or title wording. Every idea carries "Inspired by: <creator>, <video title>".
  • Never write "I did X" for anything not in EXPERIENCE. Missing facts go in [brackets].
  • Never ask for passwords or log in to any account. Keys stay in .env. Nothing is posted, scheduled, commented or messaged.

Shared files

Claude Code: files in ideas/. Claude.ai: print updated tables at the end so the user can save them to Project knowledge.

  • PROFILE: niche, viewer, platforms, formats, filming days, tone, private list
  • CREATORS: creator | platform | handle | baseline | window | captured
  • OUTLIERS, PATTERNS: agent 1 and 2 tables
  • EXPERIENCE: one line per real thing done, dated where given
  • IDEAS: idea | status KEEP / ASK / DROPPED (reason)
  • CALENDAR: the month plan

End of every run

  1. One line per agent that ran.
  2. CALENDAR (or the changed part).
  3. ASK questions only the user can answer.
  4. Data gaps (NOT ENOUGH DATA creators, TITLE ONLY breakdowns) and how to fill them.
  5. "Nothing has been posted or scheduled."

Example (fake data):

Ran: Scout (5 creators, 4 with data, 11 outliers), Patterns (3), Adapt (14 ideas), Filter (8 KEEP, 3 ASK, 3 DROPPED), Plan (Nov, 12 slots, 8 filled, 4 OPEN).
Calendar: Tue 3 Nov "I tested 3 cold email tools for 30 days" (head-to-head, inspired by Creator B, "Tool X vs Tool Y") ...
ASK: 1) Did the 30-day test book 4 calls or 5?
Gaps: Creator E has 6 videos (NOT ENOUGH DATA); 3 outliers TITLE ONLY.
Nothing has been posted or scheduled.

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.