Run any Skill in Manus with one click

google-ads-creative-inventory

Stars12

Forks5

UpdatedJuly 13, 2026 at 18:55

Map a Google Ads account's SEARCH creative into a clean, structured graph — every ad (id + name) and below it its variations (headlines and descriptions), each with its pin position, character length, Google performance label, and its own impressions/clicks/conversions/cost, plus the extensions linked to it. This is the PRODUCER of the per-account `creative_graph` that every other creative skill (performance, copy-hygiene, rewrite) consumes. Use this skill whenever the user wants to inventory / structure / map their ad copy, asks "what headlines and descriptions am I running", needs the relationship between ads and their assets, or is starting any creative audit — even if they don't say "inventory". Maps and structures ONLY; it does NOT judge performance, repetition, or copy quality (those are the consumer skills). Search/RSA first; PMax & other types are documented in references/campaign-types.md.

Installation

Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.

Run Skill in Manus

Source

portermetricsample

portermetricsample/marketing-skills

View GitHub Repository View Creator Repositories

Download

Run Skill in Manus

Related occupationsSOC

Based on SOC occupation classification

Market Research Analysts and Marketing SpecialistsBusiness and Financial Operations Occupations·SOC 13-1161

File Explorer

7 files

SKILL.md

readonly

name

google-ads-creative-inventory

description

Google Ads — Creative Inventory (the creative graph)

Goal (job-to-be-done)

Turn the account's scattered ad assets into one structured creative graph: the ad as the parent (id, name, type, ad strength, final URL, tracking), and below it its variations — headlines[] and descriptions[] — each carrying its pin position, character length, Google performance label, and its own served metrics (impressions, clicks, conversions, cost). It also attaches the extensions linked to each campaign/ad group.

The emitted JSON includes the creative_graph that the deterministic scripts of the OTHER creative skills (performance, copy-hygiene, rewrite) consume — this skill is its producer, exactly as structure-map produces the account_profile.

Who: any marketer / media buyer auditing or reporting on ad copy.
When: the first step of any creative analysis; cached per account, re-pull when ads change.
Decision it drives: nothing on its own — it gives every downstream creative skill a single, correct, structured view so they never re-pull or re-join the data.
The differentiator: it emits two views — the tree (ad → its variations: the relationships) and the rollup (each unique headline/description aggregated across every ad that uses it: the basis for performance). Structure is read from each ad's authoritative responsive_search_ad arrays (≤15 H / ≤4 D), NOT from the asset view (which over-reports); per-asset metrics are joined on by text, and each asset is tagged served (impr > 0).

Scope

✅ Map + structure Search (RSA) creative into the graph: ads, their headlines/descriptions (pin, char length, perf label, per-asset metrics), and linked extensions. Plus legacy Expanded/Text ads as a flat fallback.
✅ Optionally label segments (brand / competitor-conquest / generic) per ad group when an account_profile is supplied — so downstream benchmarks can be segment-aware. Degrades to unknown without a profile.
❌ Judge anything — winners/losers (performance), repetition/character waste/pinning (copy-hygiene), rewrites (rewrite), message→landing match (keyword-ad-landing/alignment), extension presence gaps (account-audit/ad-assets). This skill only structures.
❌ PMax / Demand Gen / Display / Video / Shopping / App creative — different asset models; Search is built first. Where each type's creative lives and what's reachable is documented in references/campaign-types.md so consumers know the boundary.

Components (read these references as needed)

Tools / data plan: references/tools.md — the 3 GAQL pulls and why this skill uses report.query (GAQL) instead of query_data (asset text is not in the reporting fields).
Framework / rubric: references/framework.md — the brain: the graph model, tree vs rollup, what's universal (code) vs profile, the served-only rule, gotchas.
Campaign-type nuances: references/campaign-types.md — where creative lives per campaign type (Search/RSA, PMax, Demand Gen, Display, Video, Shopping, App, legacy) and exactly what is reachable. Read this to understand PMax & the others.
Output schema: references/output.md — the JSON this skill emits (the creative_graph + a small synthesis).
Deterministic core: scripts/process.py + scripts/query.json — the merge/aggregate runs as code, not by the model.

Operate

Input (per account): the raw results of the GAQL pulls in references/tools.md — Q_CREATIVE (each ad's authoritative responsive_search_ad headline/description arrays + container), Q_METRICS (per-asset metrics from the asset view, used only to join by text), and Q_EXT (extensions). Optionally the account's account_profile (from structure-map) for segment labels.

Process (split brain):

Deterministic merge → scripts/process.py. Feed it the saved GAQL result files. It builds the tree from each ad's responsive_search_ad arrays (the truth — NOT the asset view, which over-reports), joins per-asset metrics by (ad_id, field, text) and tags each asset served (impr > 0), builds the rollup (aggregate each unique field+text across ads), computes char length vs the per-type limit, flags DKI, attaches extensions, and — if a profile is given — tags each ad group's segment. All universal: any account, any industry.
LLM step (thin). The model does NOT re-structure. It only writes the 3-line synthesis and, if no profile was supplied, may add a best-effort segment guess (clearly marked inferred). Apply references/framework.md.

Emit the JSON in references/output.md: synthesis (3 strings) + creative_graph (tree, rollup, extensions, coverage). Pure data — no emojis, tables, markdown, or colors. A renderer (porter-reporting) turns it into the human view.

Voice (link, don't copy): narrative lines per ../../../_framework/writing.md — plain language for a non-technical owner; name the exact ad / asset, never bare jargon.

Example (illustrative — fictional Acme Insurance, NOT rules)

Tree: ad 111111111111 ("Term Life" ad group, RSA, strength EXCELLENT, → /term-life) → 15 headlines (one pinned to position 1) + 4 descriptions, each with its 30-day impressions/clicks/conv.
Rollup: "Acme Insurance" (headline) appears in 5 ad groups → one rollup row, 10,176 impr summed, pinned_somewhere: true.
Coverage: account is 100% Search → tree complete; a PMax campaign would appear in coverage.not_mapped with reason "PMax creative lives in asset groups — see campaign-types.md".

More from this repository

same repository

porter-instagram-dashboard

portermetricsample/marketing-skills

Build a polished, hosted Instagram Insights dashboard for any Instagram account using the PorterMetrics MCP — a 4-page report (Overview, Posts, Stories, Audience) that ships as a live shareable URL. Use this skill WHENEVER the user wants an Instagram / IG analytics dashboard, report, or "insights" page built with Porter / PorterMetrics — even if they only say "make me an Instagram report", "dashboard for my IG account", "replicate this Instagram dashboard", "Porter report for Instagram", or hand you a Porter Instagram report URL to reproduce. It offers a one-call CLONE path that works anywhere (including plain chat, no build tools) plus a full build path, so every run produces a near-identical dashboard. Do NOT hand-roll an Instagram report from scratch when this skill applies.

2026-07-2412

dashboard-builder

portermetricsample/marketing-skills

Build a complete, hosted marketing dashboard on Porter Metrics end-to-end — from intake interview and SUMAS planning, through a brand-approved design kit (Porter brand or white-label extracted from any client website), to a live published report URL with real data. Use when a user with the Porter MCP asks for "a dashboard", "a client report I can share", "a live report for account X", or wants to white-label a report with a client's brand. Connector- agnostic (Google Ads, Meta, GA4, ...) and framework-agnostic (Next.js template is the default path, any static-site stack meeting the Porter contract works).

2026-07-2312

quality-score-trend

portermetricsample/marketing-skills

Track each keyword's Google Ads Quality Score over time and flag the keywords whose QS is DECLINING — the quiet erosion that raises your CPC and drops your rank without any alert firing — then attribute the drop to the component responsible (Expected CTR, Ad Relevance, or Landing Page Experience). Use this skill whenever the user asks about Quality Score trend, QS dropping / falling over time, "why did my Quality Score go down", CPC creeping up for no obvious reason, keywords slipping in rank, or which component (expected CTR / ad relevance / landing page experience) is hurting a keyword — even if they don't say "Quality Score". This skill diagnoses WHERE QS is eroding and WHY (which component) ONLY; the ad-copy fix belongs to the complementary `rsa-strength-copy-diversity-audit` skill, the landing-page fix to `landing-page-cro-audit`, and the point-in-time keyword↔ad↔landing relevance snapshot to `keyword-ad-landing-alignment`.

2026-07-1612

wasted-spend

portermetricsample/marketing-skills

Quantify the TOTAL wasted spend in a Google Ads account and pinpoint the biggest leaks — money spent on search terms, keywords, ad groups, or campaigns that returned ~0 conversions or cost far more per conversion than target. Rolls the leaks up to one dollar number and a % of spend, then lists the top leaks by scope, each with the dollars wasted and a recommended action. Use this skill whenever the user asks how much they're wasting, where the money is leaking, "what spend is doing nothing", zero-conversion keywords/terms, spend with no return, or an account-level waste roll-up — even if they don't say "wasted spend". Quantifies and prioritizes waste ONLY; writing the negatives belongs to `negative-keywords`, reallocating the freed budget to `spend-allocation`, and classifying WHY a term is irrelevant to `search-term-classifier`.

2026-07-1612

pmax-diagnostics

portermetricsample/marketing-skills

Give the advertiser the honest Performance Max read — is each PMax campaign earning its budget against the account's Search baseline (CPA/ROAS), what asset-group-level performance is actually visible, and — explicitly — what Google's API does NOT expose for PMax (search terms, placements, audience detail), so the reader trusts the numbers that ARE real. Use this skill whenever the user asks about Performance Max, PMax, "is my PMax working / worth it", PMax spend, PMax vs Search, asset groups, or the PMax black box — even if they don't say "diagnostics". Judges PMax earning-vs-baseline and the visible asset-group signal ONLY; full search-term mining and placement control are API-limited (named as gaps, not faked), and bid targets belong to the complementary `bid-strategy` skill.

2026-07-1612

google-ads-changelog

portermetricsample/marketing-skills

Build an exhaustive, human-readable Google Ads CHANGE HISTORY / changelog for a Porter-connected account — what changed, when, by whom, translated into plain English (the exact negative keyword, the budget before→after, the tCPA/tROAS, the paused campaign), grouped by day and change type, and overlaid against the daily CPA so you can see which performance moves followed which edits. Use when the user asks "what changed in my Google Ads account", "change history", "changelog", "account audit log", "who changed what", "qué cambios se hicieron en la cuenta", "historial de cambios", or wants to correlate account edits with CPA/performance. Runs on the Porter Metrics MCP (google-ads connector). One account at a time. Google retains only the last 30 days of change history.

2026-07-1612

name

google-ads-creative-inventory

description

Google Ads — Creative Inventory (the creative graph)

Goal (job-to-be-done)

Who: any marketer / media buyer auditing or reporting on ad copy.
When: the first step of any creative analysis; cached per account, re-pull when ads change.
Decision it drives: nothing on its own — it gives every downstream creative skill a single, correct, structured view so they never re-pull or re-join the data.
The differentiator: it emits two views — the tree (ad → its variations: the relationships) and the rollup (each unique headline/description aggregated across every ad that uses it: the basis for performance). Structure is read from each ad's authoritative responsive_search_ad arrays (≤15 H / ≤4 D), NOT from the asset view (which over-reports); per-asset metrics are joined on by text, and each asset is tagged served (impr > 0).

Scope

✅ Map + structure Search (RSA) creative into the graph: ads, their headlines/descriptions (pin, char length, perf label, per-asset metrics), and linked extensions. Plus legacy Expanded/Text ads as a flat fallback.
✅ Optionally label segments (brand / competitor-conquest / generic) per ad group when an account_profile is supplied — so downstream benchmarks can be segment-aware. Degrades to unknown without a profile.
❌ Judge anything — winners/losers (performance), repetition/character waste/pinning (copy-hygiene), rewrites (rewrite), message→landing match (keyword-ad-landing/alignment), extension presence gaps (account-audit/ad-assets). This skill only structures.
❌ PMax / Demand Gen / Display / Video / Shopping / App creative — different asset models; Search is built first. Where each type's creative lives and what's reachable is documented in references/campaign-types.md so consumers know the boundary.

Components (read these references as needed)

Tools / data plan: references/tools.md — the 3 GAQL pulls and why this skill uses report.query (GAQL) instead of query_data (asset text is not in the reporting fields).
Framework / rubric: references/framework.md — the brain: the graph model, tree vs rollup, what's universal (code) vs profile, the served-only rule, gotchas.
Campaign-type nuances: references/campaign-types.md — where creative lives per campaign type (Search/RSA, PMax, Demand Gen, Display, Video, Shopping, App, legacy) and exactly what is reachable. Read this to understand PMax & the others.
Output schema: references/output.md — the JSON this skill emits (the creative_graph + a small synthesis).
Deterministic core: scripts/process.py + scripts/query.json — the merge/aggregate runs as code, not by the model.

Operate

Process (split brain):

Deterministic merge → scripts/process.py. Feed it the saved GAQL result files. It builds the tree from each ad's responsive_search_ad arrays (the truth — NOT the asset view, which over-reports), joins per-asset metrics by (ad_id, field, text) and tags each asset served (impr > 0), builds the rollup (aggregate each unique field+text across ads), computes char length vs the per-type limit, flags DKI, attaches extensions, and — if a profile is given — tags each ad group's segment. All universal: any account, any industry.
LLM step (thin). The model does NOT re-structure. It only writes the 3-line synthesis and, if no profile was supplied, may add a best-effort segment guess (clearly marked inferred). Apply references/framework.md.

Voice (link, don't copy): narrative lines per ../../../_framework/writing.md — plain language for a non-technical owner; name the exact ad / asset, never bare jargon.

Example (illustrative — fictional Acme Insurance, NOT rules)

Tree: ad 111111111111 ("Term Life" ad group, RSA, strength EXCELLENT, → /term-life) → 15 headlines (one pinned to position 1) + 4 descriptions, each with its 30-day impressions/clicks/conv.
Rollup: "Acme Insurance" (headline) appears in 5 ad groups → one rollup row, 10,176 impr summed, pinned_somewhere: true.
Coverage: account is 100% Search → tree complete; a PMax campaign would appear in coverage.not_mapped with reason "PMax creative lives in asset groups — see campaign-types.md".