| name | draft-twitter |
| description | Draft a platform-native Twitter/X post or thread in your voice. Reads platform-voice.md + brand-brief.md + business-brain.md. 280-char enforcement per tweet, thread support up to 12 tweets. Plan-then-approves, writes to outputs/, optionally schedules via Blotato. |
| when_to_use | ["/draft-twitter","/draft-x","draft twitter post","twitter thread","x post about","tweet about","write a twitter post"] |
/draft-twitter — v0.1.0
Draft a Twitter/X tweet or thread. Native cowork-social skill — Blotato's post-writer doesn't cover X. Reply-driving, no hashtags, 280-char enforcement.
Inputs (lazy-load)
Read these files on every run. If any is missing, halt and tell the user which /onboard-social phase to re-run.
| File | Purpose |
|---|
about-me/writing-rules.md | Universal writing rules |
about-me/business-brain.md | ICP, audience, recent stories, wedge |
projects/social-media-content/brand-brief.md | what_you_sell, primary_cta, recent_proof_story, contrarian_belief, voice_signature |
projects/social-media-content/platform-voice.md | Twitter section only — length_range, signature_openings, signature_closings, banned_phrases, voice_notes |
projects/social-media-content/state-social.md | blotato_status (connected / skipped / awaiting-oauth) |
_shared/hook-patterns.md | 11 hook patterns (Step 4) |
Foundation A — lazy-load: read ONLY the Twitter section of platform-voice.md.
The 12-step flow
Follow the 12-step spec in _shared/draft-skill-spec.md. Platform-specific deltas below.
Platform deltas
| Field | Twitter value |
|---|
| Length per tweet | 280 chars HARD limit — validate before save, regenerate if over |
| Thread support | Up to 12 tweets via Blotato additionalPosts array |
| Image | Optional |
| Hashtags | 0-2 max (X engagement drops with hashtags — prefer 0) |
| Emojis | Sparing (≤1 per tweet) |
| First-3-words test | First 5-7 words decide whether anyone reads |
| CTA bias | Drive replies + retweets |
| Hook pattern bias | Contrarian Take (#6), Specific Number (#3), Reframe (#1) |
Thread-specific rules (insert between Step 6 and Step 7 of shared spec)
If the topic needs >280 chars, draft as a thread:
- Ask the user: "Single tweet or thread?" (default: single if topic fits)
- Thread = 3-7 tweets typical, 12 max
- Tweet 1 = hook (no preamble, no "thread incoming")
- Each tweet a complete thought — readers can land mid-thread
- Last tweet = CTA
- Validate each tweet ≤280 chars. Regenerate any that exceed.
Output location
outputs/social-media-content/YYYY-MM-DD-twitter-<slug>.md
Frontmatter:
---
date: YYYY-MM-DD
platform: twitter
slug: <kebab-case-3-5-words>
hook_pattern: <pattern-name>
length_chars: <count>
type: single | thread
tweet_count: <N>
asset_used: null
status: drafted
---
Single-tweet body
Just the tweet text. One block.
Thread body format
## Tweet 1 (hook)
<tweet text> [<char_count>/280]
## Tweet 2
<tweet text> [<char_count>/280]
...
## CTA tweet (last)
<tweet text> [<char_count>/280]
---
Type: thread (<N> tweets)
Total chars: <total>
Schedule step (Step 9-10 of shared spec)
Branch on projects/social-media-content/state-social.md → blotato_status:
connected → Load scheduling-defaults.md, find Twitter row, get next-occurrence. Ask:
"Schedule this for [today 11am ET]? (y / specific time / skip)"
On y, call mcp__claude_ai_Blotato__blotato_create_post with: text (Tweet 1), accountId (Twitter from blotato_list_accounts), optional mediaUrls, additionalPosts (array of remaining tweets if thread), scheduledFor (ISO 8601).
skipped / awaiting-oauth → Save to outputs/ only. Tell user:
"Blotato isn't connected — copy/paste the tweet(s) into X manually at [default time per scheduling-defaults.md]."
On success: update calendar-log row → scheduled, update post frontmatter → status: scheduled, return Blotato post ID.
Hard rules
- 280-char per-tweet limit enforced — regenerate any tweet that exceeds
- Plan-then-approve before any write
- Auto-grade via
/grade-post (grader scores Tweet 1 as the primary hook)
- No em dashes, no AI-throat-clearing openers
- No hashtags unless user explicitly asks for one
- Prompts to user: 3rd-4th grade reading level. The TWEET itself: voice-tuned per brand-brief + platform-voice.
Voice guardrails
Twitter post voice is defined in projects/social-media-content/platform-voice.md (Twitter section) + brand-brief.md (voice_signature). Don't reinvent it.
Self-improvement close
See _shared/foundations.md → Foundation B. After delivering the main output + the ⚡ NEXT MOVE block, ask: "What would've made this 10% better?" Log to projects/social-media-content/memory.md:
<YYYY-MM-DD> | /draft-twitter | <answer verbatim>
Run recurrence check.
Next move
See _shared/foundations.md → Foundation C. Skill output ends with a ⚡ NEXT MOVE block matching the validation regex.
Twitter NEXT MOVE priority
- Just scheduled via Blotato → "Check replies on this tweet/thread tomorrow morning + reply to top 3 within 30 minutes of seeing them"
- Just drafted + Blotato skipped → "Post the tweet manually within 2 hours of any trending topic in your niche today"
- Thread drafted → "Pin the thread for 48 hours after posting + reply to your own Tweet 1 with a 'bookmark this' CTA"
- Draft scored under 90 → "Run
/grade-post on Tweet 1 — the hook is doing 80% of the work"
Examples (the bar)
⚡ NEXT MOVE: Schedule the tweet for today 11am ET.
Why: That's your peak Twitter window per scheduling-defaults + the contrarian take rides this morning's news cycle.
⚡ NEXT MOVE: Pin the thread for 48 hours after posting tomorrow.
Why: Pinned threads compound on replies + your follower base sees Tweet 1 first when they visit your profile.
⚡ NEXT MOVE: Run `/draft-linkedin` on the same angle tonight.
Why: This contrarian take has LinkedIn-length depth + repurposing now keeps the hook timely on both platforms.
