convez.ai
SkillMarketing & VendasIntermediário
pritesh-crypto

Content Researcher

pritesh-crypto/content-researcher-skill

Transforma perfis sociais em estratégia de vídeo com relatórios data-driven

Ver no GitHub
InicialPluginScore · de 0 a 100

Skill que transforma qualquer perfil do Instagram ou TikTok em um relatório estratégico de conteúdo baseado em dados. Coleta os últimos 100 posts via Apify, executa análise estruturada em 5 fases e gera documento .docx com 28-40 páginas contendo breakdown de performance, análise de pilares de conteúdo, inteligência competitiva e um slate pronto para produção com os próximos 10 vídeos recomendados.

Instalação

$npx convez-ai pritesh-crypto/content-researcher-skill

1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.

Métricas

Estrelas
1
Forks
0
Acompanhando
0
Saúde
57%
Último commit
há 6 meses
Versão
—
Issues abertas
0
Contribuidores
1

Casos de uso

  • Fundadores D2C validarem estratégia de conteúdo com dados reais
  • Estrategistas de conteúdo identificarem pilares underutilizados e oportunidades de rebalanceamento
  • Criadores descobrirem tópicos com melhor performance antes de gravar novos vídeos
  • Times de marketing analisarem competidores e whitespace em categorias

Veredito

Skill bem construída que resolve um problema real: auditorias de conteúdo social que custam caro e entregam pouco. O output é estruturado, honesto sobre o que é dado vs. hipótese, e acionável — o destaque em 'Pattern Weight Analysis' prova valor prático. Limitação: depende de tokens Apify e Claude, não é gratuito; requer setup inicial de skills no Claude.

README

A tradução para o português deste README ainda está na fila.

Content Researcher

A Claude skill that turns any Instagram or TikTok page into a data-backed video content report.

Built for D2C founders, content strategists, and creators who want to stop guessing which Reels to shoot next. Point it at any brand's handle → get a 28–40 page .docx report with a performance breakdown, content pillar analysis, competitor deep-dive, and a production-ready slate of your next 10 videos.


Why this exists

Most content "audits" are either superficial (agency decks that say "post more Reels") or expensive (platform analytics tools that require API integrations and ongoing subscriptions).

This skill sits in between: it scrapes the last 100 posts from any public Instagram or TikTok handle, runs a structured 5-phase analysis, and outputs a report that's actually useful — one you could hand to a founder or a content lead and expect them to change what they shoot next week.

No pretending. The skill explicitly labels what's data (views, likes, pillar percentages) vs. what's hypothesis (hook effectiveness, retention risk). It does not claim to see platform retention curves — those live in the brand's own Meta Business Suite.

What you get

A structured .docx report with five sections:

  • Section A — Branded Content Strategy — performance snapshot, content pattern weight analysis, top vs. bottom video diagnosis, retention risk maps, hook strategy, pillar recommendations, 30-day direction
  • Section B — Creator / Influencer Strategy — creator archetypes, content angles, UGC vs. paid mix recommendations
  • Section C — Category & Competitive Intelligence — target vs. closest competitor comparison, pillar gaps, pillar edges, whitespace opportunities
  • Section D — The Next 10 Videos to Shoot — 3 hero briefs (full production-ready) + 7 shoot-ready concepts, ordered by confidence
  • Section E — Prioritized Action Plan — impact × effort matrix for the first 30 days

Consistently 591–623 paragraphs. Predictable, scannable, actionable.

The hero finding: Pattern Weight

In every real test, one insight has been the sharpest strategic output — the Content Pattern Weight Analysis. It calculates what percentage of the brand's posts are invested in their worst-performing content pillar, and frames that as a rebalancing opportunity.

Example from testing: a brand was spending 29% of their output on product demo reels averaging 0.75% engagement rate, while their best pillar (founder stories at 3.18% ER) was only 11% of output. That single finding was worth the whole report.

Requirements

WhatWhyCost
Claude Pro or TeamThe runtimeSubscription
Apify accountInstagram/TikTok scrapingFree tier works for small tests; ~$0.05–0.15 per report

No ML keys, no local Python, no video downloads. Just Apify + Claude.

How to use it

  1. Install the skill — clone this repo, then follow Claude's skills docs to add SKILL.md as a skill in your Claude workspace.
  2. Invoke with a natural request:
    • "Run a content audit on @brandname"
    • "Analyze @brandname's Instagram against @competitor"
    • "What kind of videos should @brandname shoot next?"
  3. Answer the briefing questions — handle, niche, region, closest competitor.
  4. Run the HTML scraper — Claude delivers a pre-filled HTML file; you paste your Apify token, click run, download the JSON.
  5. Upload the JSON back to Claude — it runs the 5-phase analysis and hands you the .docx report.

Typical end-to-end time: ~10 min of active user time, ~5 min of waiting.

Honesty rules

This skill is built on a strict set of honesty rules that are enforced throughout. PRs that weaken these will be rejected.

  • Never claim to have platform retention data. The skill has scraped outcome data (views/likes/comments) and nothing else.
  • Every Phase 3B output carries an "Analysis method: Thumbnail + caption inference" label.
  • When a competitor scrape fails (some profiles fail Apify's scraper for reasons we don't fully understand), surface it in the report rather than silently dropping them.
  • When competitor data is stale (latest post >3 months old), flag as cautionary not aspirational.

See TEST_LEARNINGS.md for the full rationale and real-world cases that shaped these rules.

Repo structure

content-researcher-skill/
├── SKILL.md                              # Main skill — load this into Claude
├── README.md                             # This file
├── CHANGELOG.md                          # Version history
├── CONTRIBUTING.md                       # How to contribute
├── TEST_LEARNINGS.md                     # Lessons from real-world testing
├── LICENSE                               # MIT
└── references/
    ├── apify-scraper.html                # Pre-filled HTML for Apify scraping
    ├── scraper-guide.md                  # Apify setup instructions
    ├── video-analysis-framework.md       # Hook taxonomy + pacing heuristics
    ├── benchmarks.md                     # Category engagement benchmarks
    ├── mock-data-schema.md               # Demo mode schema
    └── report-template.md                # Final .docx report structure

Who this is for

  • D2C founders tired of agency reports that say "post more Reels"
  • Content strategists who want a repeatable audit framework, not vibes
  • Marketers pitching creator or brand content and need data to back it
  • Anyone building on Claude who wants a reference for structured multi-phase skills

Known limitations

  • Public Instagram/TikTok accounts only (private accounts can't be scraped)
  • Apify's scraper has profile-specific failures (documented in TEST_LEARNINGS.md)
  • Per-video analysis is hypothesis-driven from caption + thumbnail metadata, not direct video frame analysis
  • Report is optimised for Indian D2C context in its benchmarks (easily adapted — see TEST_LEARNINGS.md)

Contributing

See CONTRIBUTING.md. Issues and PRs welcome, especially for niche-specific benchmark presets.

License

MIT — see LICENSE.

Credits

Built iteratively with Claude by someone who got tired of "let's post more and see what works."

Ver o original no GitHub

Bom para

Relacionados