agentleFS
aya-talebaya-taleb/video-boost-skills/skills/shorts-research/SKILL.md

shorts-research

Skill1 starsChanged 22 days agoMITView on GitHub

Researches what is working on YouTube Shorts right now for one topic, market and language - top recent Shorts (from the YouTube Data API when a key is set), their patterns, common mistakes, and YouTube's own rules (Shorts length, monetization and inauthentic content, AI disclosure, how Shorts are recommended) - using only sources from the last 12 months (6 with recent), each with a link and date. Use when the user asks what works on YouTube Shorts for a topic, why Shorts fail, or as part of viral-checklist.

  • Reads credentials

The file

We split this skill into the rules it gives an agent

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

  • 4

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  • could be automatic: a setting could enforce it
  1. Distinctive:

    Write today's date at the top of your notes.

    line 19
  2. Very specific:

    Recent top Shorts in the topic (data): If YOUTUBE_API_KEY is set, run the helper:

    line 31
  3. Very specific:

    If there is no key, search the web for recent top Shorts in the topic instead, and keep only ones with a dated view count.

    line 36
  4. Distinctive:

    Every finding needs a link and a publish date inside the window.

    line 58
  5. Very specific:

    No numbers from memory. If you cannot find a current source, leave it out or add it to gaps.

    line 60
  6. Distinctive:

    Keep claims short and specific, in plain words.

    line 61
  7. Very specific:

    When done, run source-check on the notes, unless the user asked you not to.

    line 63

≈ “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

YouTube Shorts research

Finds what is working on YouTube Shorts for one topic, and what is not, with a source and date for every finding.

Inputs

InputDefault
topicask if missing
market (country)US
languageen
window12 months (recent = 6 months)

Write today's date at the top of your notes. Every date check uses it.

Output

notes-shorts.json in the format in shared/checklist-format.md. Aim for 15 to 30 findings. List what you could not find under gaps.

Where to look, in this order

  1. YouTube's own rules and tips (official).
    • YouTube Help (support.google.com/youtube): Shorts length and format, monetization policies (inauthentic, repetitive and reused content), altered or synthetic content disclosure, fake engagement policy, made for kids rules.
    • The YouTube Official Blog (blog.youtube) and Creator Insider (YouTube's own creator channel) for how Shorts are recommended. Interviews with YouTube staff count as official when they are on a YouTube channel or quoted by a dated trade press article.
    • Find the latest version of each rule and its date. YouTube changed Shorts rules several times in 2025 and 2026.
  2. Recent top Shorts in the topic (data). If YOUTUBE_API_KEY is set, run the helper:
    python3 <skill_path>/scripts/top_shorts.py --topic "travel tips" --region US --lang en --months 12 --max 25
    It searches recent videos in the topic, keeps the ones 3 minutes or shorter, and prints title, channel, views, likes, comments, length, and post date as JSON. Each search costs 100 units of the free daily quota of 10,000. Use the results as example sources (the link is the video). Look at the top 10: the hook, length, format, text on screen, sound, call to action. Open the video page if you need to see the first seconds. If there is no key, search the web for recent top Shorts in the topic instead, and keep only ones with a dated view count. Add "no YouTube API key" to gaps.
  3. Studies and analysis (data, analysis). Recent studies with a sample size and dates (Shorts length, retention, Shorts to long-form subscribers). Mark creator surveys as surveys, not platform data.
  4. Failures (mistakes). Reused clips without changes, templated AI videos (not paid by YouTube since the July 2026 clarification; find the current page), misleading thumbnails or titles, engagement bait, wrong made for kids setting. A mistake needs a reason and a source.

What to pull out

Cover each of these, or list it as a gap:

  • Hook: first 1 to 3 seconds of top Shorts in the topic
  • Length that works now, and the current maximum length
  • Loop: does the end lead back into the start
  • Pacing, structure, payoff
  • Text on screen and captions
  • Sound: voice, music, YouTube audio library
  • Calls to action: allowed ("subscribe, like, share" is fine) vs not (incentives for engagement)
  • Shorts that lead to long videos (related video link)
  • Monetization rules: inauthentic content, AI personas on health, finance and legal topics, reused content
  • AI disclosure: when it is required (realistic content) and when it is not (clearly unrealistic, like animation)
  • Topic-specific risks

Rules

  • Every finding needs a link and a publish date inside the window. A video example uses its post date.
  • Market and language must match, or the finding is marked "scope": "global" (platform-wide announcements only).
  • No numbers from memory. If you cannot find a current source, leave it out or add it to gaps.
  • Keep claims short and specific, in plain words.

When done, run source-check on the notes, unless the user asked you not to.

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.