For Agents · MCP

Your task.
The right skill.

The tunnel uses the same cards as the catalogue. Browse only as deep as needed — or load a broad bundle immediately.

No local installationNo M11 loginRead-only

Broad task

Load 5–10 and go

SEO, Sales, Agents or another broad area → one bundle call → work.

MCP endpoint: https://skills.m11.ch/mcp

Read-only access to published skills. Default 8, maximum 10 skills / 120,000 characters.

Task Packs

Marketing Foundation

Compact two-skill starter: clarify positioning and choose a lead magnet. Use Marketing Launch for the broader eight-skill go-to-market workflow.

Marketing Launch

Build an evidence-led marketing plan from ICP and competition through positioning, campaigns, growth and measurement.

Agent Audit Essentials

Diagnose architecture and context, plan agent-team responsibilities, then organize project context and session handoffs. Memory and cost-runtime reviews remain outside this pack.

Conversion and Activation Review

Review the journey from landing page and lead capture through registration, first value and transparent upgrades.

Campaign Content and Brand Review

Plan a campaign, draft its channel content and review the work against actual brand guidance.

Content and Launch Planning

Prioritize an editorial roadmap and plan how to launch and distribute it across suitable channels.

Customer, Market and Community Research

Understand customer needs, compare competitors and plan a community around real member value.

Lead Magnet to Lifecycle Nurture

Choose a relevant lead magnet, then draft a permission-based nurture journey with entry, suppression and exit rules.

API Contract and Observability

Define the API contract, then plan how to observe its latency, failures and retries. Guidance and checklist; no production changes.

Data Analysis Foundation

Profile a dataset, choose and interpret statistical methods, then validate calculations and conclusions before sharing.

Choose an area

01 · Category

What the MCP returns at this step

84 · Marketing · Ads Operations

Google Ads Creation Workflow

Route Google Ads create/update tasks through the GoMarble proposal and approval workflow; use API Quickstart for SDK setup.

googleanzeigenanlageablaufgoogle-ads-create-master-skill

Google Ads Creation Orchestrator — GoMarble · Original SKILL.md

---
name: google-ads-create-master-skill
description: "MUST load FIRST for any Google Ads create, update, or mutation task. Master workflow for create campaigns, ad groups, ads, assets, sitelinks, experiments, negative keyword lists, bid modifiers, pause/rename/update. Triggers: any Google Ads write intent."
---
# Google Ads Mutation Workflow

Master workflow for any Google Ads create or update operation — campaigns, ad groups, ads, assets, experiments, shared negative lists, bid modifiers. Sequence is fixed: confirm intent → pull context → propose → user approves (auto-executes) → confirm.

> **Tool behavior:** All `google_ads_propose_*` tools auto-execute on user approval. There is no separate `google_ads_execute_approved_operation` call to make. When the user approves a proposal, it executes immediately — capture the returned IDs and resource names from the tool response. (`google_ads_execute_approved_operation` exists as a recovery hatch for legacy/timeout edge cases — do not use it in the normal flow.)

---

## Sub-Skill Index

Each step has a dedicated sub-skill. **Read the sub-skill at its step — not all upfront.**

| Step | Read before | Path |
|------|------------|------|
| Create / update campaign | Calling `google_ads_propose_create_campaign` or `google_ads_propose_update_campaigns` | `google-ads-create-campaign` |
| Create / update ad group | Calling `google_ads_propose_create_adgroup` or `google_ads_propose_update_adgroups` | `google-ads-create-ad-group` |
| Create / update ad (RSA) | Calling `google_ads_propose_create_ad` or `google_ads_propose_update_ads` | `google-ads-create-ad` |
| Create / update asset, update PMax asset group | Calling `google_ads_propose_create_asset`, `google_ads_propose_update_asset`, or `google_ads_propose_update_pmax_asset_group` | `google-ads-create-asset` |
| Create / update experiment | Calling `google_ads_propose_create_experiment` or `google_ads_propose_update_experiment` | `google-ads-create-experiment` |
| Create / update shared negative keyword list | Calling `google_ads_propose_create_negative_keyword_list` or `google_ads_propose_update_negative_keyword_list` | `google-ads-create-negative-keyword-list` |
| Adjust bid modifiers / ad schedule | Calling `google_ads_propose_update_bid_modifiers` | `google-ads-create-bid-modifiers` |

Always honor `google-ads-guardrails` before proposing any change.

---

## Phase 1 — Confirm the Mutation Intent

Read the user's request and identify:
- **Action** — create or update?
- **Entity type** — campaign / ad group / ad / asset / experiment / negative list / bid modifier
- **Customer ID** — must be 10 digits, no dashes. Confirm with the user only if multiple accounts are accessible or the request is ambiguous.
- **Manager ID** — required only when operating cross-account through an MCC.

If the entity type is ambiguous (e.g. user says "boost mobile by 20%" — could be DEVICE bid modifier or campaign-level mobile bid), ask exactly one clarifying question. Otherwise proceed.

---

## Phase 2 — Pull Just-Enough Context

Run targeted GAQL queries to fill in what's needed for a clean propose. Don't dump raw rows at the user — synthesize into 1–3 lines.

Common probes (run only what's relevant):
- `customer.currency_code` — needed for any budget input.
- Existing `campaign_budget.id` — when reusing a budget instead of creating one.
- `campaign.advertising_channel_type` + `campaign.bidding_strategy_type` — when switching strategies on update.
- `ad_group.id`/`name` + parent `campaign.id` — when adding ads or keywords to an existing group.
- `customer.conversion_tracking_status` — required gate for smart bidding (TARGET_CPA / MAXIMIZE_CONVERSIONS / TARGET_ROAS / MAXIMIZE_CONVERSION_VALUE).

If the account has no conversion tracking and the user asked for smart bidding, the validator will reject — instead, propose `MAXIMIZE_CLICKS` (TARGET_SPEND) and tell the user why.

---

## Phase 3 — Propose

Read the relevant sub-skill from the index above. Build the propose call with all auto-detected values filled in. Each entry in a batched call shows as its own approval row, so users can approve/decline per-item.

Output a tight summary table to the user *before* the propose tool returns its preview — keep it short:

| Field | Proposed value | Source |
|---|---|---|
| ... | ... | ... |

After the propose call returns, the agent UI shows an approval card. **Do not narrate the card.** Wait for approval/decline. Approval auto-executes the operation — the same tool response will include the execution result and any returned IDs/resource names.

---

## Phase 4 — Read the Result

The propose tool returns one merged response covering propose **and** execute. Inspect it:

- `success: true` and `execution_results[]` populated → mutation landed. Capture each entry's `entity_id`, `resource_names`, and surface the relevant IDs to the user.
- `success: false` and `execution_results[i].success === false` → mutation failed at the platform. Read `error_code`, `error`, `platform_errors[]` from that entry.
- `validation_errors[]` populated → propose-side validation rejected before any approval prompt. Fix the inputs and re-propose.
- `rejected_operation_ids[]` populated → user declined. Do nothing further unless they ask.
- `timeout_operation_ids[]` populated → user didn't act in the active window. They can still approve via the panel for 24 hours — let them know, then stop.

If a step fails:
1. Read `validation_errors[]` and `platform_errors[]` carefully. If the root cause is fixable from the agent side (e.g. validation, bad ID), correct it and retry once.
2. If the same step fails again or the error needs a manual fix in Google Ads UI (policy violation, account suspension, missing permission), stop and surface it cleanly. Don't loop on the same broken payload — explain what failed and what the user needs to do.

---

## Phase 5 — Confirm

After execution succeeds, post one short confirmation line per entity. Format:

> Done. Created campaign `<name>` (id `<id>`) ENABLED and ad group `<name>` (id `<id>`) ENABLED. Ads are PAUSED — enable when ready.

**Default status convention** (mirrors Meta):
- **Campaigns** — created `ENABLED` (active)
- **Ad groups** — created `ENABLED` (active)
- **Ads** — created `PAUSED`. Ads are the user's final gate before spend begins.

Even when the user says "launch it" or "go live", new ads stay `PAUSED` until the user explicitly approves an enable step. To enable: separate explicit propose call — `google_ads_propose_update_ads` setting `status: ENABLED`. (Approval still required; auto-executes on approval.)

We only flip between `ENABLED` and `PAUSED`. There is no removal — the API supports it, but our skill set deliberately does not. If the user asks to "delete" or "remove" an entity, propose pausing it instead and explain.

---

## Edge Cases

- **Empty account, no reference campaigns:** Use cold-start defaults — `MAXIMIZE_CLICKS` bidding, single GEO from the user's market, English language, ~$30–$50/day budget. Flag the cold start once: "No active campaigns to mirror — using defaults; happy to adjust before approval."
- **Smart bidding requested but no conversion tracking:** Propose `MAXIMIZE_CLICKS` instead. Tell the user the account needs conversion tracking before TARGET_CPA / TARGET_ROAS / MAXIMIZE_CONVERSIONS / MAXIMIZE_CONVERSION_VALUE will work.
- **Multiple items in one batch:** Use a single propose call with N items in the array — one approval card covers all, all execute together. Never split into N calls.
- **Mixed entity types in one user request** (e.g. "create a campaign with ad groups and ads"): Each entity type still goes through its own propose call. Order: campaign → capture IDs from execution result → ad group → capture IDs → ad. Same as Meta's strict order.
- **Re-using a budget (`budget_id`):** Always GAQL-confirm the budget exists and is not removed before referencing.
- **Cross-account (MCC):** `manager_id` must be set on the propose call. Verify the manager has linkage to the customer before proposing.

---

## Output Rules

- **Phase 1 / Phase 2 synthesis:** 1–3 lines. Account, currency, reference campaign or "cold start". Done.
- **Phase 3 summary table:** Compact. Only fields the user could reasonably want to override.
- **Approval card:** Don't narrate. Don't add a "shall I proceed?" message after.
- **Phase 4 errors:** Short error line + suggested fix. No raw API dumps.
- **Phase 5 confirmation:** One line per executed entity. Then stop.

Never mention "skill", "protocol", "propose tool", or internal tool names in user-facing output — describe what's happening in business terms.

When to use

Route Google Ads create/update tasks through the GoMarble proposal and approval workflow; use API Quickstart for SDK setup.

What you get

The complete original GoMarble guidance, with source attribution and declared limits.

How it works

Clarify scope, use available evidence, and apply the relevant source workflow within actual user authorization.

Requirements

Authorized account, intended resource change, currency, baseline state and connected GoMarble tools.

Delivery and review notes

Approval may execute immediately in the real GoMarble connector; check returned IDs before retries. New ads remain paused until explicit enable approval. Defaults for budgets, language or geography require visible review. GoMarble connectors, referenced sibling skills and runtime dependencies are not bundled or installed by M11. Brief obvious-danger screening only; no functional test or comprehensive safety certification. Loading this text authorizes no external calls, account mutations or ad spend.

Starting prompt

Use Google Ads Creation Workflow for [TASK]. Clarify Authorized account, intended resource change, currency, baseline state and connected GoMarble tools. Separate observed evidence from heuristics and recommendations. Check actual connector availability and approval semantics before any external action.

Not for

Claiming installed GoMarble tools, proven campaign outcomes or authorization to launch ads merely from loading this source.

What M11 added

German routing and explicit scope. Original attribution: GoMarble. Approval may execute immediately in the real GoMarble connector; check returned IDs before retries. New ads remain paused until explicit enable approval. Defaults for budgets, language or geography require visible review. GoMarble connectors, referenced sibling skills and runtime dependencies are not bundled or installed by M11. Brief obvious-danger screening only; no functional test or comprehensive safety certification. Loading this text authorizes no external calls, account mutations or ad spend.

Original authorship remains with gomarble-ai/marketing-agent · Original source ↗

Original license & copyright

MIT License

Copyright (c) 2026 GoMarble

Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the "Software"), to deal
in the Software without restriction, including without limitation the rights
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
copies of the Software, and to permit persons to whom the Software is
furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all
copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
SOFTWARE.

Source SHA-256: e6c885f220d3c827985d740b3904fc92990ff4e1b1c615c6aa8c1211aa47bbb6
Snapshot checked: 2026-09-28T16:19:52.867Z

Find a skill.Get to work.

All skills

Multi-Agent Orchestration

AI Context Window Audit

AI Agent Architecture Audit

AI Skill Discovery

Lead Magnet Strategy

Ideal Customer Profile

Go-to-Market Strategy

Marketing Campaign Ideas

Product Growth Loops

Competitor Analysis

North Star Metric

Product Positioning

Product Vision

Go-to-Market Motion Selection

Customer Feedback and JTBD Analysis

PESTLE Market Environment Analysis

Customer Journey Mapping

Ansoff Growth Options

Data Analysis Validation

Dataset Profiling

Statistical Analysis Guidance

Programmatic SEO Planning

Landing Page and Form Conversion Review

Paywall and Upgrade Planning

Signup and Registration Review

User Onboarding and Activation

Popup and Modal Planning

Email Sequence Copy and Flow

Marketing Campaign Planning

Marketing Content Drafting

Brand Voice and Content Review

Marketing Performance Reporting

Sales Company Research

Company and Contact Enrichment

Website Information Architecture

Content Strategy and Editorial Roadmap

Product Launch Planning

Customer Research and Voice of Customer

Community Growth and Member Experience

Competitor Research Profiles

Roadmap and Release Communication

Lifecycle Email Sequences

API Contracts and Interface Design

Observability and Instrumentation Planning

Agent Context and Session Handoff

Product Discovery Sprint

Deal Quality Scoring

Enrichment Waterfall Design

Customer Retention Playbook

Sales Coaching Practice

Revenue Cohort Analysis

Intent Signal Scoring

Customer Identity Matching

Segment Activation Planning

Sales Call Review

Retention Dashboard Design

Sales Coaching Competencies

Offer Delivery Design

Offer Value Communication

Service Productization

Conversion Hypothesis Prioritization

Braze Engagement Guidance

Brand Voice Guide Design

Content Pipeline Orchestration

Website Brand Profile

Buyer Objection Clarification

Klaviyo Integration Guidance

Offer Bonus Planning

Klaviyo Marketing Review

Cross-Platform Advertising Review

Expertise Business Model

LinkedIn Advertising Guidance

TikTok Advertising Guidance

Evidence Research Brief

Cinematic Video Prompts

Visual Prompt Model Adaptation

Visual Prompt Diagnostics

Short Film Development

Find a skill.
Get to work.