TikTok Data Skill
The TikTok Data Skill is the source-sliced data guide a Platform Agent owns: it names the concrete `tiktok/...` operations, response shapes, and gotchas for reading public TikTok data through UnifAPI MCP, so any creator or trend research (shortlisting, audience-fit, demand-radar, local-buzz) runs from one deterministic workflow. Read-only — it never posts.
Agent-native
Run it in Claude, ChatGPT custom MCP apps, OpenClaw, Hermes, Codex, Claude Code, Cursor, VS Code, or another MCP-capable client. No dedicated GUI flow and no separate LLM API key.
Backed by live public data
Every step is grounded in live public-data records UnifAPI returns, so the output cites what is actually ranking, posting, or being said — not a generic best-practice list.
Composable & open source
Skills cross-reference each other and live in a public, MIT-style repo. Read the full SKILL.md on GitHub, fork it, or run it as-is inside your agent.
Paste this into Codex or Claude Code
The prompt is intentionally editable. Replace the handles, market, budget, and campaign goal, then let the agent call UnifAPI MCP when it needs live public data.
Using the TikTok Data Skill, research creators for a skincare niche: search users and videos for the topic, resolve the strongest handles to ids, read their follower and video engagement, and surface rising hashtags. Return a cited brief. Read-only.
The full skill, rendered from its SKILL.md
The deterministic read path for public TikTok data through UnifAPI. This is a
Data Skill: it does not run a marketing job on its own — it names the
concrete tiktok/... operations, response shapes, and gotchas so any creator- or
trend-first workflow (shortlisting, audience-fit, demand radar, local buzz) reads
from one known recipe instead of rediscovering the surface each time.
Read-only — eyes, not hands. It researches public TikTok data and returns cited records; it never posts, comments, or follows, and UnifAPI never holds TikTok credentials.
Use the unifapi skill for live evidence
Connect once through the shared unifapi skill (OAuth MCP), then call the
operations below. TikTok users and videos are keyed by numeric id — resolve a
@handle or a video URL to its id first (see Gotchas). Keep any billing
metadata so the output can state record cost.
Response contract
Single-entity endpoints return the object in data:
{
"request_id": "unif_...",
"data": {},
"billing": { "records_charged": 1, "balance_remaining": 99 }
}
List endpoints return an array in data plus pagination:
{
"request_id": "unif_...",
"data": [],
"pagination": { "has_more": false, "next_cursor": null },
"billing": { "records_charged": 1 }
}
When pagination.has_more is true, pass pagination.next_cursor as the next
request's cursor (limit sets page size). Always preserve billing.
Core operations
| Need | Operation |
|---|---|
| Resolve handle → id | tiktok/users/resolve (?username=...) |
| Resolve video URL → id | tiktok/videos/resolve (?url=...) |
| Search videos / users | tiktok/search/videos · tiktok/search/users (?q=...) |
| Search hashtags / mixed | tiktok/search/hashtags · tiktok/search (?q=...) |
| Creator profile | tiktok/users/{id} |
| Creator videos | tiktok/users/{id}/videos |
| Followers / following / likes | tiktok/users/{id}/followers · .../following · .../likes |
| Video + reaction | tiktok/videos/{id} · tiktok/videos/{id}/comments |
| Hashtag feed | tiktok/hashtags/{id} · tiktok/hashtags/{id}/videos |
| Music feed | tiktok/music/{id} · tiktok/music/{id}/videos |
Need a field not listed here? Use the unifapi skill's get_operation to read
the exact schema before calling — but pick the operation from this table, don't
discover blind.
Workflow
The deterministic recipes. Pick the one that matches the job; each names exactly what to call.
- Discover by topic. Call
tiktok/search/videos?q=...,tiktok/search/users?q=..., ortiktok/search/hashtags?q=...for the niche; usetiktok/search?q=...for a mixed result. - Resolve a creator. Call
tiktok/users/resolve?username=...to get the numeric id, thentiktok/users/{id}forfollower_count,like_count, andvideo_count. - Read engagement. Call
tiktok/users/{id}/videos; eachVideocarriesview_count,like_count,comment_count, andshare_count— compute engagement per video. Page vianext_cursor. - Read a video and its reaction. Call
tiktok/videos/{id}(resolve a URL withtiktok/videos/resolve?url=...first), thentiktok/videos/{id}/commentsfor sentiment. - Track trends. Call
tiktok/hashtags/{id}forvideo_count/view_countandtiktok/hashtags/{id}/videos(ortiktok/music/{id}/videos) for what's spreading on a tag or sound. - Map a network. Call
tiktok/users/{id}/followers,.../following, or.../likesfor reach and affinity work. - Cite everything. Every figure ties back to the video or profile it came
from; report
billing.records_charged(or estimate when billing metadata is absent).
Shape notes
User—username,display_name,follower_count,following_count,like_count,video_count,is_verified,is_private,region.Video—video_description,view_count,like_count,comment_count,share_count,duration,author,music,hashtags,create_time.Hashtag—name,video_count,view_count,is_commerce.Music—title,author,is_original.
Gotchas
- TikTok users and videos are keyed by numeric id. Resolve a
@handlewithtiktok/users/resolve?username=...and a video URL withtiktok/videos/resolve?url=...before anytiktok/users/{id}/...ortiktok/videos/{id}/...call. tiktok/videos/batchis a POST (ids in the body), not a GET — use it to hydrate many video ids at once.- A private creator (
is_private) limits what you can read — surface it. - A low balance can silently truncate list pages: check
billing.truncated_due_to_balance— when true the page is partial, so top up before trusting any count.
Output
Return the records the calling workflow needs, each cited to its video or
profile, plus a one-line cost note (records_charged). When this skill is used
directly, a compact creator summary is the default:
**@handle** — {followers} followers, {videos} videos. Recent videos: avg {views} views / {likes} likes / {comments} comments. Rising tags: {#hashtags}. Evidence: {video URLs}. Records: ~{N}.
Related skills
- creator-shortlist, audience-fit-check (Influencer Marketing) — shortlist and vet TikTok creators on top of this read path.
- social-listening-brief (Social Listening), treatment-demand-radar, menu-demand-radar, restaurant-local-buzz (verticals) — trend and demand work that reads TikTok first.
- unifapi — the shared data skill: connect MCP and look up exact schemas with
get_operation.
Source: tiktok/SKILL.md on GitHub — open a PR there to improve it.
The live APIs this skill calls
Every operation the skill names is one of these UnifAPI platforms — still visible and callable for product code, debugging, and custom agent flows.
- The concrete `tiktok/...` operations for each common creator/trend read job
- Response-contract and shape notes for users, videos, hashtags, music
- Deterministic recipes: resolve creators, read engagement, discover trends
- Platform gotchas (resolve username to id first, cursor pagination, billing)