Draft Brand Tweet

Draft Brand Tweet

Generate tweet drafts for the @your-brand company account following 's established voice, competitor-informed positioning, and messaging guidelines.

Default: if the user asks to draft a tweet without specifying an account, draft it as @your-brand.

When to Use

• User asks to write, draft, or compose a tweet and doesn't specify an account (assume @your-brand)
• User explicitly wants a tweet for the @your-brand account
• User needs a post about product updates, , model support, partnerships, community, or events
• Use a personal-voice tweet skill only if the user explicitly asks for a specific person's personal voice

Inputs

1. Topic or objective (required)
2. Tweet type (required): feature, model-support, demo/showcase, partnership, community, event, POV/opinion, incident
3. Must-include facts/claims (optional but strongly recommended for technical posts)
4. URL/assets to include (optional)
5. Person to feature (optional — name and role, for demo/showcase tweets)
6. Length target (optional): short (80-130 chars), standard (180-270 chars)

Process

1. Validate inputs

   • If topic or tweet type is missing, ask for it.
   • If the request needs concrete claims but no facts are provided, ask for source facts or continue with clearly non-numeric/high-level wording only.

2. Pull style anchors from corpus and style notes

   • Read .agents/skills/draft-brand-tweet/style-notes.md first — this is the distilled voice guide with patterns, phrases to use/avoid, emoji conventions, tone calibration examples, and terminology guardrails.
   • Then scan your brand's recent tweet corpus (if you maintain one) for recent examples matching the tweet type.
   • Weight recent tweets highest.
   • For each draft, extract 3-5 style cues from matched corpus examples (cadence, phrasing, CTA style, length).

3. Check competitor context

   • Read .agents/skills/draft-brand-tweet/competitor-style-notes.md for competitor patterns and differentiation angles.
   • If the topic overlaps with something a competitor recently tweeted about (e.g., model support or a shared feature area), note it and ensure 's angle is differentiated.
   • Apply the "where can differentiate" insights when relevant.

4. Apply messaging and terminology guardrails

   • If you maintain a separate positioning or messaging repo, align with your own style guide and positioning docs.
   • Preserve your product naming consistently (product name, sub-products or platforms, and any capitalized feature names).

5. Ground technical claims in docs

   • Check your product docs (if available) for accuracy on technical/product claims.
   • If docs are unavailable, skip and note under "Open questions."

6. Draft candidates

   • Produce 3 options:
     • Option A: Default voice — product-first, concise
     • Option B: Shorter / punchier
     • Option C: More editorial / POV angle
   • Keep wording natural and post-ready.

7. Self-check before output

   • Style fidelity: Does this sound like the @your-brand corpus and style-notes.md?
   • Terminology: Does it match messaging docs?
   • Factual grounding: Are claims sourced? No fabricated features, metrics, or dates.
   • Banned phrases: None of the phrases from the "avoid" list in style-notes.md?
   • Competitor awareness: If relevant, is 's angle differentiated per competitor-style-notes.md?

Output

Return in this format:

1. Recommended draft
2. Alternative drafts (2)
3. Why this matches style (3-5 concise bullets)
4. Competitor context (if relevant — what competitors have said on this topic, how this differentiates)
5. Source grounding (bullet list mapping non-trivial claims to source file paths)
6. Open questions (only if required facts are missing)

Hard Constraints

• Do not fabricate product capabilities, metrics, dates, customer counts, or roadmap commitments.
• Prefer clear, plain language over hype.
• Keep tone confident and product-first, not generic "AI marketing."
• No hashtags. One emoji max.
• If uncertain, ask for missing facts or use bounded language.

Refresh Context

If you maintain tweet corpora and scraper scripts in a separate repo, refresh them there. For example:

• Refresh the @your-brand corpus: scripts/twitter-scraper/scrape_tweets.py --user your-org --since YYYY-MM-DD --cookies ...
• Refresh competitor corpora: same script with --user <competitor_a> or --user <competitor_b> and --min-likes 100
• Re-run style-anchor selection after refresh.

References

• .agents/skills/draft-brand-tweet/style-notes.md — voice and style patterns
• .agents/skills/draft-brand-tweet/competitor-style-notes.md — Competitor analysis and differentiation
• If you maintain a separate positioning or messaging repo, also check your style guide and positioning docs, plus:
  • your brand's tweet corpus (e.g. tweets/your-org.md)
  • competitor tweet corpora for differentiation (e.g. tweets/<competitor_a>.md, tweets/<competitor_b>.md)
  • any scraper tooling used to refresh those corpora