Blog-to-Social Content Pipeline

Convert Joseph Jude's blog posts into three social media formats, stored in a single reusable file per post.

When This Skill Activates

• Joseph asks to generate social media content from a blog post
• Joseph says "post {slug}" to reuse stored content
• Joseph asks to modify existing social content

Format Definitions

Format	Length	Platforms
Post	100-150 words	LinkedIn, Twitter
Tweet	≤300 chars	Bsky, Mastodon, Threads
Thread	Multi-tweet, each ≤300 chars	Twitter, Bsky

Directory Structure

All social media content (both jjude.com and CTO Field Notes) lives in ONE centralized directory. The url field in frontmatter identifies the publication.

produce/hermes/social-media/
├── content/                    # One file per blog post (both publications)
│   ├── weekly-review.md        # jjude.com
│   ├── engineering-mind.md     # CTO Field Notes
│   ├── flywheel.md
│   └── ...
├── published/
│   └── log.csv                 # Publishing tracking
├── images/
│   ├── diagrams/               # Excalidraw sources
│   └── generated/              # PNGs, prompts, style guides
├── _meta/
│   ├── templates/              # post.md, tweet.md, thread.md
│   └── prompts/                # generation-prompt.md
└── sources.csv

Do NOT create per-publication social-media directories. Keep everything in this one place.

Content File Format

Each blog post gets ONE file with all three formats:

---
title: "Blog Post Title"
url: https://jjude.com/slug/
slug: blog-slug
tier: evergreen | recent
created: YYYY-MM-DD
last_used: YYYY-MM-DD
next_use: YYYY-MM-DD


Post (LinkedIn, Twitter)

[100-150 word post]

Tweet (Bsky, Mastodon, Threads)

[≤300 char tweet]

Thread (Twitter, Bsky)

[first tweet]
[second tweet]
[third tweet]

Voice Rules (Critical)

Load the blogging skill first. All voice rules (I-voice, banned words, Flesch 65-70, ASD-STE100) come from there. This section only adds social-specific deltas:

• Pull Joseph's actual insights and frameworks — do not invent
• Conversational but professional — smart friend over coffee
• Vary hooks (story, question, counterintuitive claim, personal anecdote) to avoid formula drift

Workflow

Phase 1: Generate (first time)

1. Read the source blog post (from URL or wiki)

2. Identify the ONE sharpest insight — the idea worth sharing

3. Find a personal angle (what Joseph learned, tried, or observed)

4. Generate all 3 formats:

• post: Hook → 3-4 body paragraphs → reflection question (100-150 words)
• tweet: One sharp standalone insight (≤300 chars)
• thread: Hook tweet (standalone) → 3-4 body tweets → closing takeaway (each ≤300 chars)

5. Save as content/{slug}.md with metadata

6. Show Joseph for review before any publishing

Phase 2: Publish (reuse)

When Joseph says "post {slug}" or similar:

1. Read content/{slug}.md

2. Build variants.json for Typefully (see typefully-pipeline skill)

3. Run create_drafts.py once per format — the script creates one draft per run. For three formats (LinkedIn post, tweet, thread), create three separate spec files and run the script three times. Save specs to /opt/data/variants-{slug}-{format}.json (not /tmp/).

4. Collect the Typefully draft URLs from each run

5. Update last_used and next_use (3-6 months out) in the content file

6. Tell Joseph the drafts are ready with all URLs

Phase 3: Modify

When Joseph requests changes:

1. Read content/{slug}.md

2. Make the requested edits

3. Show Joseph the updated content

4. Save back to file

Metadata Fields

• title: Blog post title (for reference)
• url: Live blog post URL
• slug: Filename identifier
• tier: evergreen (repost anytime) or recent (time-sensitive)
• created: When the social content was first generated
• last_used: Last time this content was posted
• next_use: When to repost (3-6 months from last_used for evergreen)

Distribution Map

Format	Platforms
Post	LinkedIn, Twitter
Tweet	Bsky, Mastodon, Threads
Thread	Twitter, Bsky

Pitfalls

1. Post length creep: Posts drift toward 250+ words. Trim ruthlessly — 100-150 words is the target.

2. Thread tweet independence: Each tweet in a thread must work as a standalone quote. No "building on the previous point..."

3. Bsky char limit: Max 300 chars per post. Thread tweets must each be under 300.

4. No emoji in Excalidraw: If adding diagrams, Excalidraw's Virgil font doesn't render emoji.

5. Template drift: When generating at scale, posts can converge on a formula. Vary hooks (story, question, counterintuitive claim, personal anecdote).

6. Reuse check: Before posting, verify the content is still relevant. Don't repost outdated references.

7. Slug resolution: Blog filenames have full titles, not slugs. Read YAML frontmatter slug field to match.

8. No /tmp writes: HERMES_WRITE_SAFE_ROOT=/opt/data blocks writes to /tmp. Save temp files (like variants.json) to /opt/data/ instead, e.g. /opt/data/variants-{slug}.json. Clean up after use.

Related

• blogging skill: Load first for all voice/style/structure rules
• typefully-pipeline skill: For creating drafts via Typefully API
• excalidraw skill: For optional diagram element format reference