Write Content

Create platform-optimized content that sounds genuinely human. This skill applies all write-like-human rules automatically - every output passes through the full AI pattern detection and human writing checklist.

Before writing, always read the write-like-human skill's ai-patterns-database.md and word-tiers.md to refresh the avoidance rules.

Step 1: Gather context

Always ask:

Content type - tweet, thread, reply, quote tweet, README, bio, blog post, email?
Topic - what's the subject?
Who are you? - background, role, expertise (if not already known)
Goal - engagement, announcement, education, growth, hiring, selling?

For Twitter content, also ask:

Short, medium, or long tweet?
Reply to a specific tweet? (get the tweet text)
Your Twitter persona/voice (or link to existing tweets to match)
Sector? (dev, crypto, startup, AI, general tech, design)
Any specific angle or hot take to push?
Thread or single tweet?
Include media? (screenshot, code snippet, video)

For README content, also ask:

Project or profile README?
What does the project do? (in plain language)
Tech stack?
What makes it unique?
Installation complexity? (one-liner or multi-step)
Target audience? (devs, end users, both)

For bio content, also ask:

Which platform? (character limits vary - see bio-formulas.md)
Professional or personality-forward?
Key things to communicate (role, project, passion, CTA)
Reference bios you like?

If context is missing, ask. Don't assume.

Step 2: Write the content

Twitter content

Read tweet-templates.md for the 25 proven viral formats and hook-formulas.md for opening patterns.

Single tweets:

Pick the template that fits the content type. Generate 3 variations:

Option A (recommended) - your best shot
Option B (alternative angle) - different approach
Option C (bold/risky) - more provocative take

For each, note the expected engagement type: likes, replies, bookmarks, or retweets.

Length guidance:

71-100 chars: highest engagement rate. Hot takes, one-liners, reactions.
70-110 chars: best for retweets. Shareable, quotable.
240-259 chars: most total likes. Complete thoughts, mini-stories.
Full 280: no penalty. Use when the content needs it.

Count characters precisely. Emojis = 2 chars. Links = 23 chars always.

Threads:

Read thread-structures.md for architecture.

Structure:

Hook tweet - self-contained, creates curiosity gap. Must work even if nobody reads the thread.
Context/stakes - why should anyone care? 3-8. Body - one insight per tweet. End each with an open loop.
Key takeaway - punchy lesson.
Soft CTA - question to drive replies.
Hard CTA - follow, RT first tweet, or link (in reply).

Sweet spot is 5-10 tweets (7 is ideal). Under 5, make it a single tweet. Over 15, expect significant drop-off.

Cliffhanger every 1-2 tweets: "But that's not all..." / "Here's where it gets interesting:" / "And then everything changed."

Suggest visual breaks every 3-4 tweets (+45% completion rate).

Replies:

Not "great point!" - add genuine value:

Share your experience with the topic
Add data or a specific example
Ask a sharp follow-up question
Respectfully disagree with reasoning
3+ sentences, unique perspective
Keep niche-relevant

Quote tweets:

Add a unique take, not just agreement
"This + [your insight]" format
Expand with your experience
Or respectful disagreement with reasoning

Sector-specific tweet styles

See tweet-examples.md for examples by sector.

Dev Twitter:

Ship updates: "just shipped [feature]. here's what I learned building it:"
TIL posts: "TIL [surprising thing]. here's why it matters:"
Tool recs: "if you're not using [tool] for [task], you're spending 3x the time"
Open source: "just open-sourced [project]. it does [thing] in [impressive metric]"

Crypto/CT:

"gm. shipped a new feature overnight. wagmi."
Alpha threads: "been researching [protocol] for 2 weeks. here's what most people are missing:"
Market takes: directional, opinionated, NFA disclaimer
Community fluency: know the memes, use the slang naturally

Startup:

Build-in-public: "Month 3 update: [metrics]. Here's what went wrong:"
Hiring: "[Role] at [Company]. We're doing [exciting thing]. DM me."
Milestones: "From $0 to $[X]k MRR in [time]. No paid ads. Here's the playbook:"
Failures: "We almost died in Month 2. Here's why, and how we survived:"

README content

Read readme-templates.md for structure guides.

Project README structure:

Project name + one-line value proposition (not buzzwords - plain language)
Badges - CI, version, license, downloads (4-6 max, one row)
Hero visual - screenshot, GIF, or architecture diagram
"Why [Project]?" - the problem it solves (not "revolutionizing" anything)
Quick Start - under 60 seconds, copy-paste commands
Key Features - bullet points, specific, measurable where possible
Usage examples - real code, not pseudo-code
Installation - step-by-step
Configuration - if applicable
Contributing - welcoming tone
License - one line

Write in a human voice. The README should have personality while being informative. No corporate speak.

Profile README structure:

Opening: who you are, what you care about
Current focus: "building [X]" or "exploring [Y]"
Tech stack badges (shields.io)
GitHub stats (optional - only if impressive)
Social links
Creative element: ASCII art, code-formatted bio, custom banner

See readme-templates.md for badge syntax and widget markdown.

Bio content

Read bio-formulas.md for the 8 formulas and character limits.

Generate 3-5 variations. Count characters precisely.

Character limits:

Twitter/X: 160 chars
GitHub: 160 chars
LinkedIn headline: 220 chars
Instagram: 150 chars
TikTok: 80 chars
Discord: 190 chars
Bluesky: 256 chars

Emojis = 2 chars each. Show personality, not just credentials.

Step 3: Apply write-like-human rules

Every piece of output must pass through these checks before delivery. This is non-negotiable.

Zero Tier 1 banned words
Zero banned phrases
Em dashes: max 1 per 500 words
Sentence length varies dramatically
Contractions at 60-80% of eligible spots
Passive voice under 15%
No formulaic intro/conclusion patterns
Specific examples, not generic claims
Has personality and opinion
Reads naturally when read aloud
Adapted to the cultural context specified
No synonym cycling

If any check fails, rewrite before delivering.

Step 4: Deliver with analysis

For every content request, provide:

The content - 3 variations (recommended, alternative, bold)
Analysis - why you chose the format, expected engagement type
Post-publish strategy - what to do after posting

For Twitter specifically:

Best time to post (based on audience timezone)
Reply strategy for the first 30 minutes
Whether to put links in a reply
Suggested follow-up content to build momentum

write-content