TikTok Ad Research

• postplus-shared

  release-shell rules

• 找 TikTok 上跑得好的广告素材
• 看某个类目 / 国家 / objective 的 top ads
• 研究竞品广告怎么做 hook、卖点、CTA
• 从 Creative Center 数据里提炼 ad brief

• postplus-shared

  research preferences

• ${CLAUDE_SKILL_DIR}/references/task-shapes.md

• ${CLAUDE_SKILL_DIR}/references/input-notes.md

• ${CLAUDE_SKILL_DIR}/references/normalized-schema.md

Core Rule

• paid creative benchmarking
• hook and offer analysis
• objective / region / language comparisons
• ad-creative sourcing for briefs

• creator discovery
• community comments research
• organic content lane mapping

• ${CLAUDE_SKILL_DIR}/_postplus_shared/20-research/tiktok-research/SKILL.reference.md

Preferred Actor

• tiktok-creative-center-top-ads

Minimal Toolchain

• scrape:

  • ${CLAUDE_SKILL_DIR}/scripts/collection_actor_run.mjs

• normalize:

  • ${CLAUDE_SKILL_DIR}/scripts/normalize_tiktok_ads_dataset.mjs

• analyze:

  • ${CLAUDE_SKILL_DIR}/scripts/analyze_tiktok_ads_dataset.mjs

Release-Shell Execution Contract

• keep actor input JSON, raw datasets, normalized datasets, and analysis caches under

  <work-folder>/.postplus/tiktok-ads/

• keep only final user-facing summaries or shortlisted exports outside

  .postplus/

• start with a bounded first pass before broader ad pulls
• if hosted capability is unavailable, unauthorized, or returns a stable network error, stop immediately instead of switching to ad hoc shell glue

Recommended Workflow

1. classify the request into a paid-ad task shape
2. write a small actor input JSON
3. run the actor with a narrow scope first
4. normalize into the local ad schema
5. analyze repeated hooks, brands, objectives, regions, and CTA language
6. only then turn it into a brief or recommendation

Example

node ${CLAUDE_SKILL_DIR}/scripts/collection_actor_run.mjs \
  --collection-key tiktok-ads-top \
  --input ${CLAUDE_SKILL_DIR}/templates/top-ads-sample.json \
  --output <work-folder>/.postplus/tiktok-top-ads-raw.json

node ${CLAUDE_SKILL_DIR}/scripts/normalize_tiktok_ads_dataset.mjs \
  --input <work-folder>/.postplus/tiktok-top-ads-raw.json \
  --actor tiktok-creative-center-top-ads \
  --output <work-folder>/.postplus/tiktok-top-ads-normalized.json

node ${CLAUDE_SKILL_DIR}/scripts/analyze_tiktok_ads_dataset.mjs \
  --input <work-folder>/.postplus/tiktok-top-ads-normalized.json \
  --output <work-folder>/.postplus/tiktok-top-ads-analysis.json

Good Output

• top brands or advertisers in the sample
• dominant objectives
• repeated hook language
• repeated offer language
• repeated regions / geo scope
• CTA patterns
• duration distribution
• top ads by likes or CTR when available
• whether the sample is spotlight-curated or filter-driven

• observed ad facts
• likely creative implications
• missing evidence

Handoff

• ad video structure or spoken-line breakdown -> a dedicated visual analysis workflow
• ad creative review after humans inspect outputs ->

  skills/40-creative/creative-qa

• organic TikTok benchmark comparison ->

  skills/20-research/tiktok-research