| name | creator-analytics |
| description | Analyze a creator/account on any supported platform — profile stats, recent post performance, engagement rate, posting cadence, and top content. Use when the user says "analyze this creator/account/channel", "how is @handle doing", "engagement rate", or wants a performance summary for a profile. |
Creator Analytics
Profile a single account and summarize its performance. For comparing multiple accounts, use
competitor-analysis.
Setup gate
[ -z "${TIKHUB_API_KEY:-}" ] && echo "Set TIKHUB_API_KEY first (see tikhub-onboarding)."
Workflow
- Identify platform + handle from the user's input (URL or @handle).
- Resolve the user to the id the API needs (e.g. TikTok/Douyin
sec_user_id via
handler_user_profile; Twitter screen_name/rest_id; Instagram username). See the
platform skill for the exact endpoint.
- Fetch the profile (followers, following, total likes/posts, bio) — the user-info endpoint.
- Fetch recent posts (e.g. last 30–100) via the user's post-list endpoint, paginating with
the platform's cursor. Cap the page count and tell the user the cost.
- Compute metrics from the posts:
- Engagement rate ≈ avg(likes + comments + shares) / followers.
- Posting cadence (posts/week from timestamps).
- Top 5 posts by engagement; median vs. top performance.
- Trend over time (rising/declining views).
- Deliver a concise report: headline stats, engagement rate, cadence, top content, and 2–3
observations.
Endpoint pointers (per platform)
Use tikhub-find-endpoint "user info" --platform <slug> and "user posts" --platform <slug>, or
the platform skill: tiktok, douyin, instagram, youtube, twitter-threads, xiaohongshu.
Cost awareness
Profile = 1 call; each page of posts = 1 call. Estimate before running:
GET /api/v1/tikhub/user/calculate_price?endpoint=<path>&request_per_day=<n>. Warn the user
before pulling many pages.
Verification gate
- Profile resolved (non-empty follower/post counts).
- Post list non-empty and timestamps parse.
- Engagement math sanity-checked (rates between 0–100%).
Red flags
- Reporting an engagement rate from too few posts — note the sample size.
- Forgetting to resolve
sec_user_id/numeric id before the post-list call.