agentleFS
slilbudgetslilbudget/knowledge-delta-skills/skills/meta/meta-ads/SKILL.md

meta-ads

Skill3 starsChanged 49 days agoMITView on GitHub

Diagnoses and scales Meta/Instagram ad accounts: 'why is CPA up', 'CBO vs ABO', 'pixel vs CAPI mismatch', 'ad rejected', 'what does Marketing API error X mean'. Covers ODAX, Advantage+, attribution, policy, unit economics, canonical API error catalog. Not API launch/grey infra (meta-grey-ops), cross-account kill/scale (senior-buyer-ops), tracker math (tracker-ops) or causal test validity (measurement-experimentation-ops).

28 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use linkedin-ads, google-analytics and 3 more.

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-devSkill

    linkedin-ads

    LinkedIn advertising strategy for B2B lead generation, brand awareness, and recruitment. Use when running LinkedIn ad campaigns.

    Also in 4 projects that use this

  • refly-aiSkill

    google-analytics

    Integrate with Google Analytics for web metrics. Use when you need to: (1) retrieve website analytics reports, (2) track traffic and user metrics, or (3) analyze web performance data.

    Also in 3 projects that use this

  • huiferSkill

    local-seo

    本地 SEO 分析和优化专家。自动检测项目是否需要本地 SEO,分析 NAP(Name, Address, Phone)一致性、本地关键词优化、Google Business Profile (GBP) 优化和本地结构化数据生成。提供本地商家搜索引擎排名优化建议,包括 NAP 标准化、本地关键词策略、GBP 完整性检查、评论策略、地图嵌入和本地 SEO 审计。

    Also in 3 projects that use this

  • indranilbanerjeeSkill

    keyword-research

    Research keywords via the connected data source: intent, gaps. Grouping → keyword-cluster. \"what keywords should we target\"

    Also in 3 projects that use this

  • AnthropicSkill

    onboarding

    Generate an onboarding checklist and first-week plan for a new hire. Use when someone has a start date coming up, building the pre-start task list (accounts, equipment, buddy), scheduling Day 1 and Week 1, or setting 30/60/90-day goals for a new team member.

    Also in 3 projects that use this

The file

We split this skill into the rules it gives an agent

9 rules. 0 are common, 7 are distinctive, and 2 are very specific to this project.

  • 0

    Common

    Most of the 21 edited versions we read keep these.

  • 7

    Distinctive

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Read only what the task needs:

    line 15
  2. Distinctive:

    Always read 00 for current policy, eligibility, numerical claims, or external benchmarks.

    line 35
  3. Very specific:

    Do not assume the retired location-presence selector exists.

    line 43
  4. Very specific:

    Treat support replies as account-specific evidence, not platform contracts (00 §2).

    line 52
  5. Distinctive:

    Do not delete/rebuild assets or profiles to evade enforcement.

    line 55
  6. Distinctive:

    Verify authentic ownership, least privilege, 2FA, billing, policy/category.

    line 93
  7. Distinctive:

    Use distinct concepts with native 4:5 and 9:16 assets.

    line 99
  8. Distinctive:

    Before activation verify Account Quality, billing, identity, schedule, currency, budget, destination, previews, and effective statuses.

    line 102
  9. Distinctive:

    No spend: inspect restrictions, billing, status, schedule, eligibility, audience/placement, bid, and creative.

    line 118

“Common” is measured: kept by at least 6 in 10 of the 21 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.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Meta Ads

Operate as a senior Meta Ads practitioner. Treat UI, eligibility, policy, API behavior, prices, benchmarks as volatile — verify current primary sources when the answer depends on current behavior. Research reviewed per file (see each header); only sections marked verified 2026-09-2x were rechecked against primary sources. Baseline 2026-09-03 (Sonnet 5 blind, 4 tasks): 3 [W] (attribution/targeting fields, enhancement opt-outs, error codes), 1 [A]; rerun with skills fixed 5/5 flagged items.

Route references

Read only what the task needs:

NeedReference
Evidence labels, source priority, refreshreferences/00-evidence-and-maintenance.md
Portfolio, assets, access, billing, securityreferences/01-business-portfolio-setup.md
Ads Manager UI, metrics, reports, rulesreferences/02-ads-manager-interface.md
Objectives, structure, CBO/ABO, special categoriesreferences/03-campaign-objectives-structure.md
Placements, formats, ratios, creativereferences/04-instagram-placements-creatives.md
Advantage+ and manual audiences, geo, retargetingreferences/05-targeting-audiences.md
Budgets, bidding, learning, scaling, costsreferences/06-budgets-bidding-costs.md
Policy, restricted verticals, review, appealsreferences/07-policies-restricted-niches.md
Pixel/dataset, CAPI, events, UTMs, attributionreferences/08-tracking-pixel-optimization.md
Intake, economics, diagnosis, test designreferences/09-diagnostics-unit-economics.md
Sourced cases and post-mortemsreferences/10-practical-case-library.md
Budget/funnel playbooks by business modelreferences/11-budget-and-funnel-playbooks.md
Creative metrics, audits, exports, scenariosreferences/12-creative-diagnostics-account-audits.md
API/MCP governance, identity, billing, launch gates (token/scope/MCP mechanics live in meta-grey-ops/02; launch tooling in meta-grey-ops/scripts)references/13-api-access-billing-launch-operations.md
API errors: code → cause → fixreferences/14-api-error-catalog.md

Always read 00 for current policy, eligibility, numerical claims, or external benchmarks. Read 13 before API/MCP automation, accepting tokens, billing work, restrictions, or activation.

Guardrails

Tripwires that must fire before any reference is read:

  • Do not assume the retired location-presence selector exists; Special Ad Audiences are discontinued (full retired-concepts table → 00 §4).
  • Pixel is the web data source; datasets group events (08). Interests/lookalikes and Advantage+ availability vary — verify the live flow (05).
  • 50 events/7 days, budget-change percentages, frequency caps, and refresh cadences are heuristics, not universal rules. Hard limits consolidated in 00 §4.2.
  • Card verification, balance, payment eligibility, and restriction are separate states, as are portfolio membership, ownership, and creation quota. Leave does not equal deletion; "warmed" assets guarantee neither approval nor payment trust (01, 13).
  • Treat support replies as account-specific evidence, not platform contracts (00 §2).
  • System User tokens are bearer secrets. Never expose them in chat, URLs, screenshots, repositories, or logs.
  • Do not delete/rebuild assets or profiles to evade enforcement.

Evidence rules

Label material claims when useful: official behavior, Meta-reported result, independent benchmark, practitioner heuristic, or unverified. Preserve source population, geography, objective, date, methodology. Never convert a benchmark or case lift into a platform rule or forecast.

Minimum context

Infer first; ask only for missing inputs that change the decision:

  • country, currency, vertical, special-category status;
  • business model, offer, destination, sales cycle;
  • objective, conversion location/event, attribution;
  • spend/date range, structure, delivery state, recent edits;
  • CPM, CTR, CPC, LPV, CVR, CPA/CPL, value/ROAS, frequency;
  • backend revenue, margin, lead quality, refunds, lag;
  • Pixel/dataset, CAPI, UTMs, consent, Diagnostics.

State assumptions and continue when possible.

Model

Portfolio -> assets/access/billing/security
  Campaign -> objective/category/budget
    Ad set  -> event/audience/placements/bid/schedule/attribution
      Ad    -> identity/creative/copy/CTA/destination

Diagnose at the level owning the setting. Check campaign, ad-set, and ad effective status separately.

Workflow

  1. Derive break-even CPA/CPL/ROAS from margin, value, and close rate.
  2. Verify authentic ownership, least privilege, 2FA, billing, policy/category.
  3. Choose the closest measurable business outcome and conversion location.
  4. Validate one event per real outcome, parameters, CAPI deduplication, UTMs.
  5. Consolidate enough to learn; avoid redundant low-spend ad sets.
  6. Choose Advantage+ audience/placements deliberately; use available hard controls and manual structure only for a reason.
  7. Use distinct concepts with native 4:5 and 9:16 assets; preview placements and enhancements.
  8. Predefine KPI, lag, observation window, stop/scale/rollback conditions.
  9. Before activation verify Account Quality, billing, identity, schedule, currency, budget, destination, previews, and effective statuses.
  10. Reconcile Meta with analytics/CRM/backend before optimization conclusions.

For API launch, verify scopes and exact asset tasks, then create every object PAUSED as a zero-spend write probe. A successful GET does not prove write access. (A PAUSED ad created via API is still submitted for review at once; the grey-ops launcher creates ACTIVE by default, meta-grey-ops/00.)

Diagnose in order

eligibility/billing -> delivery -> auction -> attention -> click quality
-> landing continuity -> conversion -> business value -> attribution
  • No spend: inspect restrictions, billing, status, schedule, eligibility, audience/placement, bid, and creative.
  • High CPA: decompose CPM -> CTR -> click-to-LPV -> CVR -> value/quality.
  • Meta/backend mismatch: validate count/deduplication, then attribution, lag, consent, modeled/view-through results, refunds, and qualification.
  • Creative: use reference 12; define hook/hold denominators before diagnosis.
  • Export audit: request stable IDs, raw counts, daily/placement exports, attribution context, and backend joins; run scripts/analyze_ads_export.py.
  • Restriction: capture exact affected asset/reason, correct it, then appeal. (Operator lane, own BM: on an ad disapproval do not edit or resubmit; disabled account → meta-grey-ops/23.)
  • API read-only failure: inspect scopes, System User asset tasks, app/business relationship, identity, and restriction state.

Output

Lead with the decision. Include evidence level/assumptions, exact actions at the correct level, measurement window, stop/rollback conditions, and remaining uncertainty. Use cases as mechanism analogies, not outcome promises.

For regulated or sensitive verticals, verify current Meta policy and local law. Cloaking, replacement accounts, and review-layer filters are Circumventing systems. This skill does not own that lane — operational grey execution is meta-grey-ops.

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.