TikTok Research

Research public TikTok evidence for the user's marketing decision. A product, category, problem, competitor, account, or content URL is enough to start. Use the task reference below and shared contract for research completion, reporting, and downstream handoffs.

Reference Index

User asks for	Apply
Brand account, competitor account, profile health, account comparison	references/account-and-competitor-audit.md
Organic examples, hooks, hashtags, formats, competitor videos, related videos	references/organic-benchmark.md
Comments, audience voice, objections, FAQ, phrase bank, buyer language	references/audience-voice.md
Creators, UGC makers, KOL/KOC, profile enrichment, creator shortlist	references/creator-discovery.md
Paid ads, paid hooks, CTA, offers, region/objective ad examples	references/paid-ads.md
Product demo fit, ecommerce angle, buyer objection, content-to-offer fit	references/product-content-fit.md
Launch campaign, hashtag challenge, branded activity, competitor campaign	references/campaign-scout.md
Market, region, language, cross-border localization, local angle comparison	references/market-localization-scout.md
Profile facts only	Run a bounded tiktok-profiles or tiktok-users lookup
Mixed paid and organic	Run separated paid and organic lanes; never fill one lane with the other
Shop, LIVE, private analytics, backend audience, GMV, conversion, hidden contacts, exact targeting, spend, ROAS	Stop or ask for a supported public TikTok scope
Music/archive download or audio extraction	Not provided on the current public surface. Say so and stop; analyze user-provided local files through media-analysis instead
Cross-platform request	Run only the TikTok lane here; hand off other platforms

Understand the request

Infer the intended decision and useful deliverable from the request. Read an unfamiliar product's official page before deriving brands, use cases, problem phrases, and hashtags. Verify brand/account identity instead of guessing from a name. Do not ask the user to supply competitor lists or choose search batches.

Ask one focused question only when missing product identity, subject, market, or a public/private boundary would materially change the work. If the user has provided a usable product/category/account/URL, start; do not ask for another seed or an output-format choice. Preserve all platforms the user requested.

Keep explicit requirements, inferred search directions, and unknown conditions separate. Inferred directions are not hidden exclusion rules. Unsupported private metrics remain unknown; use the available public evidence only when it can still answer the requested decision.

Public Command Boundary

Run the selected route with its declared semantic flags:

postplus research run <route> --<query/url/handle/hashtag flags> --limit <n> --wait --output result.json

Evidence need	Route	Inputs
Organic videos	tiktok-videos	query, hashtag, handle, or URL flags
Known profiles	tiktok-profiles	repeat --handle
Account discovery	tiktok-users	repeat --query
Related-video discovery	tiktok-related-videos	repeat --url, after selecting relevant seed videos
Actual comments	tiktok-comments	repeat --url; optional --include-replies
Paid examples	tiktok-ads-top	--limit only; no keyword/brand/market targeting

Use postplus research run <route> --help when flags, enums, or bounds are unclear; inspect postplus research schema --route <route> --json only when processing needs exact result fields. postplus doctor --skill tiktok-research is a diagnostic, not a compulsory preflight. Never use private supplier inputs. Save independent batches separately; run them concurrently only after their inputs are known. Follow the returned structured action for pending/failed operations; resume the same operation, not a replacement submission.

postplus research run tiktok-ads-top \
  --wait \
  --output ./result.json

Follow the CLI's structured result and reported next action; do not infer recovery from free-text messages. Reuse explicit user approval that already covers the requested scope and maximum cost; otherwise obtain approval before confirming a quote. Never infer spending authorization from a task alone. Resume the same operation through its returned checkpoint or action; never resubmit uncertain work, repeat exhausted recovery, or switch providers to bypass failure.