AI Content Repurposer#

Name & Purpose#

Takes one long-form piece of content (article, blog post, video transcript, podcast episode) and intelligently repurposes it into multiple platform-optimised formats: Twitter threads, LinkedIn posts, newsletter editions, and short video scripts. Eliminates the "create once, post once" waste.

Prerequisites#

RequirementVersion/DetailOpenClawv2.4+Python 3v3.10+FFmpegv4.4+ (for video excerpts)NLP librariesSee requirements.txtAPI Key (optional)ElevenLabs for voiceover scripts

Installation#

1. Copy skill files#

bash

cp -r streams/01_ClawHub_Skills/02_AI_Content_Repurposer/* ~/.openclaw/skills/

2. Install dependencies#

bash

cd ~/.openclaw/skills/ai-content-repurposer
pip install -r requirements.txt

3. Configure#

Edit config/repurpose.yaml with your brand voice preferences:

yaml

brand:
  name: "YourBrand"
  voice: "professional-but-warm"
  industry: "tech"
  target_audience: "SaaS founders and marketers"
  tone_adjectives: ["confident", "insightful", "approachable"]
  hashtags:
    default: ["#YourBrand", "#Industry"]
    twitter: ["#YourBrand"]
    linkedin: ["#YourBrand", "#IndustryInsights"]
  cta:
    default: "Want more? Subscribe to our newsletter."
    newsletter: "Loved this? Get more insights delivered weekly."

output_formats:
  twitter_thread:
    enabled: true
    max_tweets: 15
    max_chars_per_tweet: 280
    include_cta: true
  linkedin_post:
    enabled: true
    max_chars: 3000
    include_hashtags: true
    include_cta: true
  newsletter:
    enabled: true
    max_chars: 5000
    include_header: true
    include_footer: true
  video_script:
    enabled: true
    max_duration_seconds: 60
    include_timestamps: true
    format: "tiktok"  # Options: tiktok, youtube_shorts, instagram_reels
  tweet:
    enabled: true
    max_chars: 280
    count: 5

Usage#

Repurpose a single piece of content#

bash

# From a URL (article, blog, transcript)
openclaw skill run ai-content-repurposer --input https://example.com/blog-post

# From a local file
openclaw skill run ai-content-repurposer --input ./my-article.md

# From a YouTube video (auto-transcribe)
openclaw skill run ai-content-repurposer --input https://youtube.com/watch?v=XXXX

Specify output platforms#

bash

# All platforms
openclaw skill run ai-content-repurposer --input article.md --output all

# Specific platforms
openclaw skill run ai-content-repurposer --input article.md --output twitter,newsletter

# Just video scripts
openclaw skill run ai-content-repurposer --input transcript.md --output video_script

Batch process#

bash

openclaw skill run ai-content-repurposer --batch ./content_queue/ --output all

Workflow#

text

Source Content (article/transcript/video)
  → Extract Core Narrative (key points, quotes, data, angle)
  → Platform Router
     ├── Twitter Thread: Hook → Tweets (numbered) → CTA
     ├── LinkedIn Post: Headline → Body (with line breaks) → Hashtags
     ├── Newsletter: Subject → Intro → Sections → Outro
     └── Video Script: Hook → Scene breakdown → Voiceover text
  → Quality Checks (character limits, tone alignment, quote accuracy)
  → Output Files (saved to ./output/{platform}/)

Available Commands#

CommandDescription/statusShow queue depth, recent jobs, completion stats/queueList items waiting for processing/outputsShow recent output files/schedule <cron>Set recurring schedule (e.g., /schedule "0 8 * * 1-5")/configShow current repurpose configuration/preview <job_id>Preview outputs before publishing

Repurposing Templates#

Templates define how each platform's output is structured. Edit ./templates/.

text

templates/
├── twitter_thread/
│   └── default.yaml
├── linkedin/
│   ├── thought_leader.yaml
│   ├── company_page.yaml
│   └── short_tip.yaml
├── newsletter/
│   ├── weekly_digest.yaml
│   └── deep_dive.yaml
├── video_script/
│   ├── tiktok.yaml
│   ├── youtube_shorts.yaml
│   └── instagram_reels.yaml
└── quality_checks.yaml

Twitter Thread Template (default.yaml)#

yaml

structure:
  hook:
    type: "question_or_controversial_opener"
    length: "1 tweet"
    rules:
      - "Start with a bold claim or surprising question"
      - "Hook must fit in 280 chars"
  body:
    type: "numbered_tweets"
    count: "8-15"
    rules:
      - "Each tweet = 1 idea maximum"
      - "Include line breaks within tweets for readability"
      - "Number tweets: 1/10, 2/10, etc."
      - "Quote exact source where needed"
  cta:
    type: "final_tweet"
    rules:
      - "Summarise or give opinion"
      - "Add CTA: follow, retweet, subscribe"
      - "Include 2-3 relevant hashtags"
  spacing:
    type: "smart_delay"
    config:
      initial_delay_minutes: 0
      tweet_interval_minutes: 3
      thread_label: true

LinkedIn Post Template (thought_leader.yaml)#

yaml

structure:
  headline:
    type: "bold_claim"
    length: "1-3 lines"
    rules:
      - "Lead with the contrarian take or key insight"
      - "Use all caps for emphasis on 2-3 words max"
  body:
    type: "story_problem_solution"
    sections:
      - name: "context"
        length: "2-4 lines"
      - name: "problem"
        length: "3-5 lines"
      - name: "insight"
        length: "4-6 lines"
      - name: "actionable_takeaway"
        length: "3-5 lines"
    rules:
      - "Keep paragraphs to 2-3 lines max (mobile readability)"
      - "Use bullet points for lists"
      - "Include 1-2 line breaks between sections"
  engagement:
    type: "question"
    rules:
      - "End with an open-ended question to drive comments"
  hashtags:
    count: "3-5"
    rules:
      - "Mix broad (#Marketing) and niche (#SAFounders) tags"

Quality Checks#

The system runs automated quality checks on every output before saving:

yaml

# templates/quality_checks.yaml
checks:
  - name: character_limits
    description: "Enforce platform-specific length limits"
    severity: error
  - name: tone_alignment
    description: "Check output matches brand voice configuration"
    severity: warning
    threshold: 0.7 # confidence score
  - name: quote_accuracy
    description: "Ensure quotes from source are exact"
    severity: error
  - name: link_integrity
    description: "Verify all links are from trusted domains"
    severity: warning
  - name: duplicate_content
    description: "Flag near-duplicate sentences across outputs"
    severity: warning
  - name: hashtag_relevance
    description: "Check hashtags match content topic"
    severity: warning
  - name: cta_presence
    description: "Every output must have a CTA"
    severity: error

Input Format Spec#

The system accepts these input formats:

FormatExtensionNotesMarkdown.mdPreferred. Headers = sectionsPlain text.txtAuto-detect paragraphsHTML.htmlStripped and convertedYouTube URLhttps://...Auto-transcribe via subtitlesPodcast RSShttps://...Fetch and transcribe latest episodePDF.pdfText extraction (via PyMuPDF)WordPress URLhttps://...Fetch and parse article content

Scheduling#

Use cron-style scheduling for regular repurposing:

bash

# Run every Monday at 8 AM on the latest blog post
openclaw cron add "0 8 * * 1" "ai-content-repurposer --input https://blog.yoursite.com/latest --output all"

# Auto-repurpose RSS feed items as they drop
openclaw skill run ai-content-repurposer --watch-rss https://blog.yoursite.com/feed.xml

Example Prompts for Human Operators#

"Hey Marvis, repurpose this article into a Twitter thread and a LinkedIn post: [URL]" "What's in the queue for today's repurposing schedule?" "The last TikTok script was too long — tighten it to 30 seconds max." "Run quality checks on all pending outputs." "Schedule this newsletter draft for Friday at 10 AM."

Directory Structure#

text

ai-content-repurposer/
├── SKILL.md
├── README.md
├── config/
│   ├── repurpose.yaml        # Brand voice & platform config
│   └── sources.yaml          # RSS feeds, WordPress URLs to watch
├── templates/
│   ├── twitter_thread/
│   │   └── default.yaml
│   ├── linkedin/
│   │   ├── thought_leader.yaml
│   │   ├── company_page.yaml
│   │   └── short_tip.yaml
│   ├── newsletter/
│   │   ├── weekly_digest.yaml
│   │   └── deep_dive.yaml
│   ├── video_script/
│   │   ├── tiktok.yaml
│   │   ├── youtube_shorts.yaml
│   │   └── instagram_reels.yaml
│   └── quality_checks.yaml
├── prompts/
│   ├── extract_narrative.yaml
│   ├── twitter_thread.yaml
│   ├── linkedin_post.yaml
│   ├── newsletter.yaml
│   ├── video_script.yaml
│   └── quality_check.yaml
├── output/
│   ├── twitter/              # Generated Twitter threads
│   ├── linkedin/             # Generated LinkedIn posts
│   ├── newsletter/           # Generated newsletter editions
│   └── video/                # Generated video scripts + timestamps
├── scripts/
│   ├── batch.sh              # Batch process directory
│   ├── schedule.sh           # Set up cron jobs
│   └── publish.sh            # Auto-post to social platforms (opt-in)
├── requirements.txt
└── package.json

Troubleshooting#

SymptomLikely CauseFix"Content too short"Source under 500 wordsAdjust min_length in repurpose.yamlTwitter threads > 280 charsTemplate token budget wrongLower max_chars_per_tweetVideo scripts too wordyDuration section misconfiguredCheck max_duration_seconds in templateQuotes off by one wordTranscription errorFix source, re-run with --forceNewsletter missing sectionsArticle has no clear structureAdd Markdown headers to sourceBatch job hangsOne large file blocking queueRun --max-size 10000 to splitTone sounds offBrand voice not configuredUpdate config/repurpose.yaml adjectives