google-ads-experiments

google-ads-experiments is a skill for Claude Code, Codex from chanktb/claude-google-ads. It costs 106 tokens per session (837 once invoked), scanned A, original, MIT.

A tool for designing controlled A/B tests in Google Ads. An A/B test compares two versions while changing one planned factor to see which performs better.

In plain words
What is it for?
Use it to test bidding, brand exclusions, new-customer settings, campaign structure, ad copy, audiences, or landing pages with feasibility checks and defined success rules.
Why use it?
It prevents unclear tests that change several things at once or end before there is enough data to support a decision.

Skill for Claude CodeCodex

Installs and runs on its own, but its text points at files inside its plugin — anything it tells you to read at a ${CLAUDE_PLUGIN_ROOT} path is only there once the plugin is installed. Installing the plugin gets both.

Part of the claude-google-ads plugin — 15 skills, 15 commands shipped together

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add skills/chanktb/claude-google-ads/experiments
Any agent
npx skills add chanktb/claude-google-ads --skill experiments
Clone the repo
git clone --depth 1 https://github.com/chanktb/claude-google-ads

Made for: Claude Code, Codex.

Or install claude-google-ads, the plugin that ships this one along with the rest of its 15 skills, 15 commands.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for google-ads-experiments

README.md
[![agentmods](https://agentmods.dev/badge/skills/chanktb/claude-google-ads/experiments.svg)](https://agentmods.dev/skills/chanktb/claude-google-ads/experiments)
Your own site
<a href="https://agentmods.dev/skills/chanktb/claude-google-ads/experiments"><img src="https://agentmods.dev/badge/skills/chanktb/claude-google-ads/experiments.svg" alt="Measured on agentmods" height="20"></a>
Per session 106 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 837 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00106 $0.00837
Opus 5 $0.00053 $0.00418
Sonnet 5 $0.00021 $0.00167
Haiku 4.5 $0.00011 $0.00084

Measured 5d ago against content hash 18f3f13775b4, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

google-ads-experiments scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 5d ago.

The scan reads SKILL.md. This mod also ships 1 executable file (scripts/significance.py), listed below but not scanned — reading those needs a real analyzer, not pattern matching.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

skills/experiments/SKILL.md · 56 lines

How it starts

The opening of the file, as written. The whole thing — 56 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Turn a "let's test X" into a clean experiment: one variable, enough volume to learn something, a defined readout. This is how the split menus in the builders get validated — not by guessing.

Model dispatch (run cheap, decide expensive) — see ${CLAUDE_PLUGIN_ROOT}/references/model-tier-dispatch.md

  • Scout (haiku) — STEP 2 significance.py run.
  • Routine (sonnet) — pulling conversions-per-arm estimates / recent volume for the feasibility check. Dispatch as general-purpose; return raw numbers.
  • Judge (main session) — STEP 1 variable choice, STEP 3 design + decision rule, STEP 4 platform mapping, STEP 5 readout call (against the pre-stated rule, not early noise). The power math is mechanical; designing a test that can conclude is judgment.

STEP 1 — Pick ONE variable

From the catalog (campaign-level: bidding strategy, tROAS level, brand-exclusion on/off, new-customer mode, split structure A vs B from ${CLAUDE_PLUGIN_ROOT}/references/pmax-split-strategies.md; ad-group/asset-group-level: copy theme, audience signal, landing page). Change exactly one — never bundle variables (you won't know what moved).

STEP 2 — Feasibility check (don't run a test that can't conclude)

Using ${CLAUDE_PLUGIN_ROOT}/references/forecasting-and-benchmarks.md, estimate conversions per arm over the planned runtime. If each arm won't accumulate enough conversions to detect a meaningful difference, say so and either: extend the runtime, increase budget, pick a higher-volume variable, or skip the test. A test too small to reach significance is worse than no test — it invites false conclusions. Runnable check: python ${CLAUDE_PLUGIN_ROOT}/skills/experiments/scripts/significance.py --conv-per-arm N [--mde 0.15] → smallest detectable lift + whether the run is powered for your target effect.

STEP 3 — Design

  • Control vs variant, 50/50 split, one variable.
  • Primary success metric (e.g. ROAS, CPA, conversions) + guardrail metrics (don't win on CPA while tanking volume).
  • Minimum runtime (cover learning + at least 2-4 weeks; avoid mid-experiment changes).
  • Pre-state the decision rule: what result ships the variant, what reverts it.

Read the full file on GitHub · 56 lines

Files

What ships with it

1 file beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 5d ago First seen · 56 lines · 106 tokens per session scan A 18f3f13775b4

Subscribe to this mod's changes

google-ads-experiments is a skill published in the GitHub repository chanktb/claude-google-ads (11 stars, last pushed 1mo ago), licensed MIT. It adds 106 tokens to every session and 837 once invoked, about $0.0005 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.

Related

Other skills, from other repositories

google-ads

Спецификации Google Ads для российского рынка 2026 — 8 форматов (RSA / Performance Max / Display / Demand Gen / Video / App / Call / Shopping), полный набор character limits (headline 30, description 90, long-headline 90, business-name 25 и т.д.), creative-frameworks под RSA-логику динамической сборки, Google Ads…

VKirill/claude-lane-stack · 373 tokens

markifact-overview

Reference — what Markifact is, what the MCP server exposes, and the discover→inspect→run pattern. Always loaded into the performance-marketer agent.

markifact/markifact-mcp · 36 tokens

safe-write-operations

Reference — rules for safely executing write/destructive operations against ad accounts. Always loaded into the performance-marketer agent.

markifact/markifact-mcp · 28 tokens

google-ads-ecommerce

Build and optimize Google Ads campaigns for ecommerce with Performance Max, Shopping feeds, conversion tracking, and Smart Bidding strategies for ROAS.

finsilabs/awesome-ecommerce-skills · 32 tokens

setup-conversion-tracking

Set up Converly conversion tracking end to end. Use when the user wants to track form submissions as conversions, send leads to an ad platform like Google Ads, Meta or Google Analytics, or set up Converly for the first time.

converlyio/converly-mcp · 55 tokens

diagnose-missing-conversions

Find out why Converly conversions aren't showing up. Use when the user says tracking isn't working, conversions are missing, a form submission didn't appear, or an ad platform isn't receiving conversions.

converlyio/converly-mcp · 45 tokens