tiktok-ads
Skill3 starsChanged 49 days agoMITView on GitHub
TikTok Ads strategy and diagnosis: research-first intake, objectives and optimization events, account structure, targeting, bidding and budgets, creative and Spark Ads, tracking and attribution design, policy and restricted verticals, catalogs and TikTok Shop. Executing launches via MCP/API = tiktok-ops.
10 other projects use a version of this.
Adopted by
Most recently taken up by
- adkit/ads-skillstheir own version, first seen Nov 15, 2026
- penyy8540-bit/tiktok-ads-agent-kittheir own version, first seen Nov 15, 2026
- gomarble-ai/marketing-agenttheir own version, first seen Oct 8, 2026
- vidual-labs/digital-marketing-skillstheir own version, first seen Oct 2, 2026
- slilbudget/knowledge-delta-skillsword for word, first seen Sep 28, 2026
- ruslan2027/adako-mcptheir own version, first seen Sep 28, 2026
Used with
Projects with it also use linkedin-ads.
Other files people use alongside this one.
Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
brainbytes-devSkilllinkedin-ads
LinkedIn advertising strategy for B2B lead generation, brand awareness, and recruitment. Use when running LinkedIn ad campaigns.
Also in 3 projects that use this
Other versions10 projects use a version of this.
Word for word · 1
- slilbudget/knowledge-delta-skills★ 3100% same
Edited · 0
None found.
Wrote their own version · 9
- hyperfx-ai/marketing-skills★ 112same name
- adkit/ads-skills★ 40same name
- gomarble-ai/marketing-agent★ 15same name
- vidual-labs/digital-marketing-skills★ 3same name
- penyy8540-bit/tiktok-ads-agent-kit★ 2same name
- ruslan2027/adako-mcp★ 1same name
- windsor-ai/claude-windsor-ai-plugin★ 1same name
- angelgabrieljacintohuayllasco/marketing-skills★ 0same name
- gmmh1/claude-paid-media-skills★ 0same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository17 other files it gives its agents.
- measurement-experimentation-opsSkill
- senior-buyer-opsSkill
- tracker-opsSkill
- deep-researchSkill
- knowledge-delta-skill-architectSkill
- web-scrapingSkill
- design-stack-pickerSkill
- normcore-webSkill
- responsive-adapterSkill
- google-adsSkill
- google-feed-opsSkill
- google-grey-opsSkill
- meta-adsSkill
- meta-grey-opsSkill
- js-obfuscationSkill
- secure-codingSkill
- tiktok-opsSkill
The file
We split this skill into the rules it gives an agent
5 rules. 0 are common, 2 are distinctive, and 3 are very specific to this project.
0
Common
Most of the 7 edited versions we read keep these.
2
Distinctive
This author's own choices. Worth a look.
3
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 86
Read only what the task needs.
- Very specific: line 105
Always read 00 before stating a policy, an eligibility, a limit, or a benchmark.
- Very specific: line 106
Always read 08 before touching a regulated vertical — before any ad object exists.
- Distinctive: line 129
Use cases and benchmarks as mechanism analogies, never as forecasts.
- Very specific: line 138
Before any MCP create call, tiktok-ops runs ttops preflight (validates the plan and emits the exact tool arguments with operation_status: …
“Common” is measured: kept by at least 6 in 10 of the 7 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
TikTok Ads
Reviewed 2026-09-14 against TikTok's full official documentation plus RU/EN practitioner
sources. UI, eligibility, policy, pricing and benchmarks are volatile — verify the current primary
source when the answer depends on current behaviour (00).
Operate as a senior practitioner who has not worked this account before: derive, do not assume.
This skill ships no default strategy — on purpose
The three inputs that decide structure — the payout event, the conversion volume the account can feed, and the GEO's policy and auction — vary more than the platform does. Anything shaped like a universal launch template is an artefact of one GEO and one vertical.
Read references/01-research-protocol.md before proposing anything. It owns the intake, the
constraints gate, the bounded research pass, and the derivation. The rest of this skill is the
material that protocol draws on.
Diagnosing an existing account rather than planning a new one? Go straight to 09 (diagnosis order)
and the reference the symptom names — but come back to 01 § H before you change the strategy,
because "which input was wrong" is a derivation question, not a reporting one.
intake → constraints gate → bounded research → derive → pre-mortem → plan → execute → reviewGuardrails — these fire before any reference is read
- TikTok's
operation_statusdefaults toENABLE. An object created without it is live and spending. Opposite of Meta. bid_typehas exactly two values —BID_TYPE_NO_BID(Maximum Delivery) andBID_TYPE_CUSTOM(Cost Cap). Meta's bid vocabulary does not exist. Value/ROAS bidding is a separate axis (deep_bid_type) and cannot be combined with Cost Cap.- TikTok publishes two learning numbers and they disagree — "~25 results or 7 days" before
volatility falls, and "7 days to reach 50 conversions" for an underperforming ad group. Use 25
to decide when a number is readable and 50 to decide whether a structure can be fed (
05). Porting Meta's hard 50-conversion exit gate makes you over-consolidate and mis-time every verdict. - The "±20% budget change resets learning" rule appears in no TikTok document. The one documented budget rule is that an update must be ≥105% of current spend.
- There are no location exclusions. You positively select. Rewrite any exclusion-based plan.
- Attribution window locks at ad-group creation and is not a reporting parameter. Two windows means two ad groups, decided up front.
- Currency, timezone and country are permanent at ad-account creation, and budget minimums are
base × per-currency ratio. A USD template on a JPY (×100) or IDR (×10,000) account is wrong by orders of magnitude, upward. - There is no
SALESobjective enum (it isWEB_CONVERSIONS/PRODUCT_SALES+virtual_objective_type), no Store Traffic objective, and Collection Ads can no longer be created. - Tracking fails silently. Hash mismatches,
event_idmismatches and droppedttclidthrow no error anywhere. Fire one real test conversion and watch it arrive before launching. - A rejected ad cannot be enabled into life — build a new one. And an appeal re-reviews the whole ad group, with one appeal allowed.
Minimum context
Infer what you can; ask only for what changes the decision, and state your assumptions.
GEO · vertical · offer and which tracker status pays · conversion lag · break-even CPA/ROAS · expected daily conversions at target CPA · account type and your role on it · tracking stack · creative supply per week · budget and decision horizon · account currency and timezone.
Missing the payout event, break-even, or tracking → you cannot judge a result.
Missing GEO, vertical, or access → you cannot know the campaign is legal to run. 01 § A.
Model
Business Center -> members / partners / assets / billing
Ad account -> currency + timezone (permanent), balance, role
Campaign -> objective, special categories, CBO
Ad group -> optimization goal, event, bid, budget, placements, targeting, ATTRIBUTION (locked)
Ad -> identity, creative, text, CTA, destinationDiagnose at the level that owns the setting. Campaign, ad group and ad each carry their own status — check all three separately.
Route references
Read only what the task needs.
| Need | Reference |
| Evidence labels, source precedence, doc conflicts, what expires | references/00-evidence-and-maintenance.md |
| The research-first workflow — read before proposing anything | references/01-research-protocol.md |
| BC, roles, asset sharing, agency accounts, billing, suspension deadlines | references/02-account-model-and-access.md |
| Objectives, goal×billing matrix, Smart+, GMV Max, structure | references/03-objectives-and-structure.md |
| Targeting fields, custom audiences, lookalike limits and GEO gate | references/04-targeting-and-audiences.md |
| Bidding, VBO unlock thresholds, budgets and currency math, learning, pacing | references/05-bidding-budgets-delivery.md |
| Creative specs, safe zones, identity types, Spark Ads and expiry, rejections | references/06-creatives-identity-spark.md |
| Pixel, Events API, dedup, attribution, EMQ, trackers and macros, consent | references/07-tracking-pixel-events-api.md |
| Policy, restricted verticals, per-GEO gates, review and appeals | references/08-policy-verticals-geo.md |
| Diagnosis order, fatigue, benchmarks, scaling, reporting mechanics | references/09-diagnostics-and-optimization.md |
| Catalogs, TikTok Shop, GMV Max exclusivity, commerce reporting | references/10-ecommerce-catalog-shop.md |
| RU/EN field knowledge, agency cabinets, transcript extraction, evidence quality | references/11-practitioner-field-notes.md |
| Vertical shapes — templates for the derivation, not strategies | playbooks/ |
| Execution: official TikTok MCP server (the default surface), preflight/audit guard rails, error and status codes | skill tiktok-ops |
Always read 00 before stating a policy, an eligibility, a limit, or a benchmark.
Always read 08 before touching a regulated vertical — before any ad object exists.
Do not port Meta
Universal media-buying principles do port: break-even math, sample size, one-variable tests, cohorting on click date, the operating contract. Platform mechanics do not.
Ported assumptions that are wrong here: objects create paused · bid-strategy names · a hard
50-conversion learning exit gate (TikTok has no exit gate; it publishes 25 and 50 for different
questions, 05) · ±20% budget rule · location exclusions · attribution as a reporting setting ·
Collection Ads · Store Traffic · silent-autoplay creative (TikTok requires audio as policy) ·
"3–5 ad groups × 3–5 creatives, ABO then CBO" as doctrine (11 — it is Meta convention with no
TikTok test behind it).
Where a rule came from Meta practice, say so and check it against a TikTok source before it reaches a plan.
Output
Lead with the decision. Then: evidence level and assumptions · exact actions at the correct level · budgets in major units with the currency named · the measurement window and attribution basis · stop/scale/rollback conditions · what you did not verify.
Use cases and benchmarks as mechanism analogies, never as forecasts. If you cannot name a number's source and date, do not state the number.
Boundary
Strategy and diagnosis live here. Execution lives in tiktok-ops, whose default surface is the
official TikTok MCP server — it covers essentially every operation, needs no API key, and the
operator authorizes it once in a browser. Nothing in this skill authorizes spend.
Before any MCP create call, tiktok-ops runs ttops preflight (validates the plan and emits the
exact tool arguments with operation_status: DISABLE); after it, ttops audit (diffs the read-back
and flags anything live). Both run offline with no token.
For regulated or sensitive verticals, verify current TikTok policy and local law (08). This skill
documents the compliant path and what it costs; it does not implement cloaking, review-layer
filtering, fabricated events, or landing-page swaps after approval.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

