Brand Content Strategy

Turn a website / personal brand into a research-backed content + channel strategy: what to post, where, for whom, and a 30-day plan to get known and generate leads. It is an orchestration skill — it drives reel-discovery, the perplexity_ask MCP, and the avatar/audio pipeline (viral-video-script, voice-clone, avatar-reel-composer, and the audio-theater skill) — and produces one markdown deliverable.

When to use

• "How do I get my site / brand known?" / "position me as an expert."
• "What kind of videos should I make? Who should I target?"
• "Build me a YouTube / TikTok / content strategy for <domain>."
• "Research what's working in my niche and turn it into a plan."

Inputs to gather (ask only what's missing)

Setup

• YouTube (recommended): set YT_API_KEY for exact counts — see reel-discovery.
• TikTok/IG/Facebook (optional, robust): set APIFY_TOKEN.
• perplexity_ask MCP for the creator footprint research (step 5).

Workflow

- [ ] 1. Profile the brand   (fetch the site → positioning, credentials, assets)
- [ ] 2. Design niches+geos  (3-6 queries: core niche, broad authority, news, per language)
- [ ] 3. Discover            (reel-discovery per query, --sort velocity)
- [ ] 4. Filter to humans    (scripts/filter_creators.py → drop brands, rank people)
- [ ] 5. Reverse-engineer    (perplexity_ask each top creator: footprint + funnel)
- [ ] 6. Synthesize          (positioning, platforms, pillars, 30-day plan, funnel, audio)
- [ ] 7. Write deliverable   (<brand>-30dias.md from the template in REFERENCE.md)

1. Profile the brand

Fetch the site (WebFetch). Extract the brand's positioning, credentials, proof points, products/services, and unfair advantages — these become the credential shortcut (see Principles). Also audit the site as the funnel destination: funnel coverage (top/mid/bottom), the existing CTA, locales/languages, and whether analytics exist — this seeds the website punch list (step 6).

2. Design niches + geos

Derive 3–6 discovery queries from the brand's niche. Cover, at minimum:

• the core niche (the brand's real differentiator),
• one broad authority term,
• one news / trend angle,
• the same core niche per target language (EN for US/CA, ES for LatAm…).

3. Discover

Run reel-discovery per query, sorted by velocity (what's breaking out now):

python3 .cursor/skills/reel-discovery/scripts/discover.py \
  --topic "<query>" --platforms youtube,tiktok \
  --sort velocity --since 120 --max-duration 240 \
  --limit 25 --per-platform 12 --region US --lang en \
  --slug <slug>

Repeat per language with --region/--lang (e.g. --region CL --lang es). Read each discovery/<slug>/results.md for the "Cómo están publicados" section (hashtags, sounds, channel sizes, captions) — that is the packaging playbook.

4. Filter to humans (exclude official/brand accounts)

Official product accounts (Claude, Kimi, Unity, monday…) are not competition. Collapse all discoveries into a ranked humans-only shortlist:

python3 .cursor/skills/brand-content-strategy/scripts/filter_creators.py \
  --slugs <slug1>,<slug2> \
  --keywords "agent,claude,coding,ia,automat,llm,cursor,codex" \
  --min-views 8000 --sort velocity --top 30 \
  --out discovery/_creators.md

It drops a default brand stoplist (extend with --exclude), ranks individuals by views/day, and extracts links found in their descriptions (first read on their footprint). Rank by velocity, not raw views — raw views favor mega media channels and old hits.

5. Reverse-engineer the playbook

For the top ~6 individual creators (and the ES cluster), research their full distribution with perplexity_ask. Use the prompt template in REFERENCE.md → "Creator footprint prompt". Capture per creator: platforms + followers, what they monetize, signature hook/format, and how they repurpose across platforms. Flag unverifiable claims.

6. Synthesize the strategy

Combine the brand profile + publishing patterns + creator playbooks into: positioning, audience + platform priority, content pillars, hook/bio formulas, a 30-day calendar, the funnel + monetization ladder, a website punch list (P0/P1/P2 — always include it; see REFERENCE.md §9), metrics, the production pipeline, and (optional) the audio/podcast layer. Apply the Principles below.

7. Write the deliverable

Write <brand>-30dias.md using the full template in REFERENCE.md → "30-day strategy template".

Principles (the reverse-engineered playbook)

Apply all — they are how the winners actually grow:

1. Credential in the first second / bio (authority shortcut). Lead with the brand's strongest proof, not the product.
2. Hook = result + number + tool, never the product first (e.g. "How I built [outcome in $/time] with [tool/agents]").
3. Funnel architecture: short-form (TikTok/Reels/Shorts = discovery) → long-form (YouTube = proof + SEO) → link-in-bio → site.
4. One asset → many surfaces. Produce once, distribute to all (and clip long-form into shorts). This repo's pipeline is that engine.
5. Monetization ladder: free lead magnet/workshop → course/cohort → community → high-ticket consulting (the real back-end).
6. Borrowed authority: guest on existing podcasts / press.
7. Language arbitrage: an underserved language/region (e.g. Chile/LatAm in Spanish) is far easier to own than the saturated EN market.

Anti-patterns

1. Ranking brands as competitors — always run filter_creators.py first.
2. Ranking by raw views for trend-spotting — use velocity.
3. Pure AI-voice impersonation in a credibility brand — clone the person's own voice (voice-clone) and declare it; never fake a real guest's voice.
4. Fully automated publishing with no human-in-the-loop — quality > volume.
5. Inventing formats — steal proven ones (see viral-video-script).
6. Treating the podcast as a discovery channel — it deepens/converts; clips and short-form do the discovery.

Outputs

• discovery/<slug>/ — raw reel-discovery research (gitignored).
• discovery/_creators.md + .json — humans-only ranked shortlist.
• <brand>-30dias.md — the strategy deliverable.

Utility scripts

Additional resources

• Brand-profiling checklist, query-design heuristics, the perplexity footprint prompt, the full distribution playbook, the 30-day template, and the audio / podcast layer: REFERENCE.md