/social-presence

Social Presence

Research what works, find content opportunities from your recent work, and craft high-performing tweets. Delegates to /tweet for final polish and screenshot wrapping.

Gotchas

• Don't fabricate metrics -- never claim "this got 500 likes" without verification. Use patterns observed from successful accounts, not invented numbers.
• Voice preservation -- the user has a natural dev voice. Never corporate-ify it. "Excited to announce" is banned. So are hashtags unless explicitly requested.
• Visual-first platform -- every tweet recommendation must include a visual suggestion. Text-only tweets from dev accounts get buried.
• Don't over-thread -- single tweets outperform threads for most content. Only suggest threads for major launches with multiple features.
• Release notes are not tweets -- never dump a changelog into a tweet. Extract the one thing a reader cares about and lead with that.
• Ecosystem tagging -- always suggest relevant accounts to tag (@nuxt_js, @vuejs, @unjsio, @AnthropicAI, etc.) but max 2 per tweet. More looks desperate.
• Time zones -- user is in Australia. Best posting times for global dev audience: 8-10am US Eastern = 11pm-1am AEST. Suggest scheduling if timing matters.

Data Storage

# Log content ideas that were used
echo "$(date -I) | MODE | TOPIC | FINAL_TWEET_SUMMARY" >> "${CLAUDE_PLUGIN_DATA}/social-presence.log"

Read prior logs to avoid repeating angles and to track what content types the user gravitates toward.

Input

$ARGUMENTS determines the mode:

• Release/launch context (mentions release, version, changelog, launch) -> Launch Mode
• "ideas" / "what should I tweet" / empty -> Discovery Mode
• Specific topic ("tweet about unhead", "post about this PR") -> Craft Mode

Mode 1: Discovery ("What should I tweet about?")

Find tweetable moments from recent work. Run these in parallel:

1a. Scan Recent Activity

# Recent merged PRs across all repos (last 7 days)
gh search prs --author=@me --merged --sort=updated --limit=20 -- "merged:>=$(date -d '7 days ago' +%Y-%m-%d)" 2>/dev/null || gh search prs --author=@me --merged --sort=updated --limit=20

# Recent releases
gh search releases --owner=harlan-zw --sort=created --limit=10 2>/dev/null

# Recent commits on key projects
for repo in harlan-zw/nuxt-seo harlan-zw/unhead harlan-zw/unlighthouse; do
  echo "=== $repo ==="
  gh api "repos/$repo/commits?per_page=5&since=$(date -d '7 days ago' +%Y-%m-%dT00:00:00Z)" --jq '.[].commit.message' 2>/dev/null
done

1b. Check What's Trending

Use WebSearch to check:

• What's trending in the Vue/Nuxt/UnJS ecosystem this week
• Any recent discourse where the user can add expertise (SEO, head management, Nuxt modules)

1c. Score and Rank

For each potential topic, score on:

Present the top 3-5 ideas as a ranked table (fewer if the week was quiet; don't pad with weak moments):

| # | Topic | Hook | Visual | Why now |

For each idea, include:

• A draft hook (first line of the tweet)
• What visual to create (screenshot, demo GIF, code snippet)
• Why this would perform well

Ask the user which to develop. Then proceed to Craft Mode.

Mode 2: Launch ("I'm releasing X")

For major releases, version bumps, new projects. The goal: maximize reach for a single moment.

2a. Understand the Release

Read whatever the user provides (release notes, changelog, PR list, repo). Identify:

• The single most impressive/useful thing in this release
• Who benefits and why they should care
• What's visually demonstrable

2b. Research Comparable Launches

Use WebSearch to find 2-3 recent successful launch tweets from accounts in the same ecosystem. Analyze:

• What hook did they use?
• What visual format worked? (screenshot, video, demo GIF, before/after)
• Did they thread or single-tweet?
• What engagement did they get?

Check these accounts for recent launch patterns:

• @antfu7 (Anthony Fu, Vitest/UnoCSS/Slidev)
• @danielroe (Daniel Roe, Nuxt)
• @haydenbleasel (Hayden Bleasel, next-forge)
• @shadcn (shadcn/ui)
• @tailwindcss (Tailwind CSS)

2c. Generate Launch Strategy

Present 3 tweet approaches with different angles:

1. Problem-first: Lead with the pain point this solves. "Your Nuxt site's meta tags are probably wrong. Here's a free module that fixes all of them automatically."
2. Show-don't-tell: Lead with the visual/demo. "Watch: full site SEO audit in 30 seconds." + screenshot/video
3. Milestone/story: Lead with the journey. "6 months ago Nuxt SEO had 200 stars. Today v3 ships with [headline feature]."

For each approach include:

• Full draft tweet text (under 280 chars)
• Visual recommendation (what to screenshot/record)
• Suggested accounts to tag (max 2)
• Whether to thread (usually no)
• Best time to post

2d. Visual Planning

Based on chosen approach, suggest specific visuals:

• Code snippet: use the /tweet skill's screenshot wrapper for terminal/editor screenshots
• Before/after: split image showing the problem vs the solution
• Demo GIF/video: record a 15-30 second screencast of the headline feature
• Stats image: GitHub stars, downloads, performance benchmarks

If the user has a screenshot or wants one wrapped, hand off to /tweet for the screenshot wrapping step.

Mode 3: Craft ("Tweet about X")

For specific topics the user wants to post about.

3a. Research the Topic

Quickly check:

• Has this topic been tweeted about recently by others? What angle did they take?
• Is there a trending conversation to hook into?
• What's the most interesting angle only this user can take?

3b. Apply the Playbook

Reference references/playbook.md for content rules. Key principles:

• Lead with value to the reader, not what you did
• Frame around the reader's problem
• Tag ecosystem accounts for amplification

3c. Generate Variations

Produce 3 variations:

1. Value-first: Frames around what the reader gets ("Free tool that does X")
2. Insight: Shares a non-obvious takeaway ("I assumed X, turns out Y")
3. Build-in-public: Shows the human side ("Just shipped X after wrestling with Y for a week")

For each show:

• Tweet text
• Character count
• Visual suggestion
• Why this angle works

3d. Hand Off

Once the user picks a direction, offer to:

• Invoke /tweet for final polish and screenshot wrapping
• Schedule suggestions (best posting times)
• Plan a follow-up tweet for the next day (engagement compounds with consecutive posts)

Cross-Cutting Rules

Strip AI Tells

Before finalizing any copy, run it through /humanize-writing. For social specifically: kill the em-dashes and "it's not X, it's Y" pattern, name specific tools/numbers instead of vague claims, and address the reader directly. Tweets that sound AI-written get ignored.

The "So What?" Test

Every tweet must pass: "If I saw this in my timeline and didn't follow this person, would I care?" If not, reframe around the reader's benefit.

Content Mix Targets

Over any 2-week period, aim for this mix:

• 50% project updates and launches (the core)
• 25% technical insights and tips
• 15% build-in-public / personal journey
• 10% community engagement (sharing others' work with your take)

Consistency Over Virality

A mediocre tweet posted is better than a perfect tweet drafted. The goal is 4+ posts per week. Suggest the user batch-create content on weekends for the coming week.

If no visual is obvious, default to a code snippet screenshot.