agentleFS
claude-worldclaude-world/claude-agent/.claude/skills/content-creator/SKILL.md

content-creator

Skill5 starsChanged 8 months agoMITView on GitHub

Create algorithm-optimized social media content (Threads, Instagram, etc.) from a topic or trend. Uses Meta's patent-based scoring (5 dimensions) to maximize reach. Optionally generates NotebookLM image cards. Use when user says "create a post about", "write content for", "make a post", "turn this into a post".

21 other projects use a version of this.

Adopted by

Most recently taken up by

The file

Show its rules and text

We split this skill into the rules it gives an agent

19 rules. 0 are common, 12 are distinctive, and 7 are very specific to this project.

  • 0

    Common

    Most of the 13 edited versions we read keep these.

  • 12

    Distinctive

    This author's own choices. Worth a look.

  • 7

    Very specific

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

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

    Check for with-image flag

    line 44
  2. Distinctive:

    Fallback: If trend-pulse unavailable, skip this step and write based on general best practices.

    line 52
  3. Distinctive:

    Never write from titles alone: Always read at least one primary source.

    line 55
  4. Very specific:

    If input is a URL: call browser_markdown(url=<URL>) to read the full page

    line 57
  5. Very specific:

    If input is a keyword: use WebSearch to find the top result, then WebFetch or browser_markdown to read it

    line 58
  6. Distinctive:

    Fallback: If cf-browser unavailable, use WebFetch.

    line 60
  7. Distinctive:

    Fallback: Use known defaults if unavailable.

    line 68
  8. Distinctive:

    Write the post content respecting platform character limits.

    line 81
  9. Very specific:

    Call get_scoring_guide() and score your draft against the 5 dimensions (each 0-100).

    line 84
  10. Very specific:

    Call get_review_checklist() and verify:

    line 88
  11. Distinctive:

    No AI filler phrases

    line 89
  12. Very specific:

    If with-image flag is set and notebooklm is available:

    line 95
  13. Very specific:

    If notebooklm unavailable: skip and note "Add with-image after setting up NotebookLM (uvx notebooklm login)."

    line 102
  14. Distinctive:

    Remove URLs from the post body (keep the text description)

    line 131
  15. Distinctive:

    After publishing the main post, immediately reply to it with the links:

    line 133
  16. Distinctive:

    Always read sources before writing — no exceptions

    line 153
  17. Distinctive:

    NEVER put URLs in the Threads/IG post body — always use link comment

    line 155
  18. Distinctive:

    Remove AI filler: "in today's rapidly evolving...", "it's worth noting...", "dive deep into..."

    line 157
  19. Distinctive:

    Keep the reply concise — full details go to the workspace file

    line 159

“Common” is measured: kept by at least 6 in 10 of the 13 edited versions of this file we could read. “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

Content Creator

You are the content creation skill for Claude-Agent. Create publication-ready, algorithm-optimized social media content.

Input

$ARGUMENTS — topic/URL, optional platform (default: threads), optional with-image flag.

Steps

1. Parse Input

  • Extract topic or URL
  • Extract platform (default: threads)
  • Check for with-image flag

2. Get Content Brief

Call get_content_brief(topic=<topic>) for:

  • Hook examples (5 categories: curiosity, controversy, story, data, engagement)
  • Patent strategies (EdgeRank, Andromeda, Dear Algo, etc.)
  • CTA templates

Fallback: If trend-pulse unavailable, skip this step and write based on general best practices.

3. Read Source Material (MANDATORY)

Never write from titles alone. Always read at least one primary source.

  • If input is a URL: call browser_markdown(url=<URL>) to read the full page
  • If input is a keyword: use WebSearch to find the top result, then WebFetch or browser_markdown to read it

Fallback: If cf-browser unavailable, use WebFetch. If that also fails, note the limitation.

4. Get Platform Specs

Call get_platform_specs(platform=<platform>) for:

  • Character limits (Threads: 500, Instagram: 2200)
  • Algorithm priorities
  • Best posting times

Fallback: Use known defaults if unavailable.

5. Write Draft

Apply the 5-dimension patent framework:

DimensionWeightTechnique
Hook Power25%First line: number or contrast, 10-45 chars
Engagement Trigger25%CTA anyone can answer, direct "you" address
Conversation Durability20%Present both sides, invite debate
Velocity Potential15%Timely, 50-300 chars total, shareable
Format Score15%Line breaks, mobile-scannable

Write the post content respecting platform character limits.

6. Self-Score

Call get_scoring_guide() and score your draft against the 5 dimensions (each 0-100). Calculate weighted total. Must reach >= 70. If below, revise and re-score.

7. Review Checklist

Call get_review_checklist() and verify:

  • No AI filler phrases
  • Timeline words are accurate
  • Source is cited or referenced
  • CTA is clear

8. Generate Image Card (if requested)

If with-image flag is set and notebooklm is available:

  1. nlm_create_notebook(title=<topic>) — create notebook
  2. nlm_add_source(notebook_id, text=<post_content + source_material>) — add content
  3. nlm_generate(notebook_id, type="slides", lang=<user_lang>) — generate visual
  4. nlm_download(notebook_id, type="slides", path="workspace/card-[slug].pdf") — save

If notebooklm unavailable: skip and note "Add with-image after setting up NotebookLM (uvx notebooklm login)."

9. Save

Write to workspace/content-[slug]-[date].md:

# Content: [Topic]
Platform: [platform]
Score: [total]/100
Date: YYYY-MM-DD

## Post
[The actual post content]

## Score Breakdown
- Hook Power: [score]/100
- Engagement Trigger: [score]/100
- Conversation Durability: [score]/100
- Velocity Potential: [score]/100
- Format Score: [score]/100

## Image
[Path to image card if generated, or "N/A"]

## Source
[URL or reference used]

10. Link Comment (IMPORTANT for Threads/Instagram)

Threads and Instagram do NOT support clickable links in post body. If the content references source URLs:

  1. Remove URLs from the post body (keep the text description)
  2. Collect all source URLs into a "link comment"
  3. After publishing the main post, immediately reply to it with the links:
    Sources:
    - [Title] URL1
    - [Title] URL2
  4. Save the link comment content in the workspace file under ## Link Comment

This is critical for Threads — posts with URLs in the body get reduced distribution.

11. Present to User

Show:

  • The ready-to-post content (without URLs in body)
  • Link comment content (URLs to post as first reply)
  • Score with breakdown
  • Image path (if generated)
  • Suggested posting time based on platform specs
  • Offer: "Ready to post? Want me to adjust anything?"

Rules

  • Always read sources before writing — no exceptions
  • Score must be >= 70 before presenting
  • NEVER put URLs in the Threads/IG post body — always use link comment
  • Conversation Durability must be >= 55 individually
  • Remove AI filler: "in today's rapidly evolving...", "it's worth noting...", "dive deep into..."
  • Match user's language (check profile for preferred language)
  • Keep the reply concise — full details go to the workspace file

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.