ads-test — Claude Code skill: paid-ad A/B tests
Walks a paid-media experiment from hypothesis to readout: randomization unit, sample size, stopping rule, and a plain-language decision.
npx skills add AgriciDaniel/claude-ads --skill ads-test -a claude-codeWhat ads-test does
The skill takes a paid-ad test through its steps in order. First it pins down the decision, the causal hypothesis, treatment and control, the randomization unit, the population, the primary metric, guardrails, the minimum effect and the stopping rule. Then it checks what the ad platform allows: overlapping tests, conversion lag, seasonality, cross-group interference and whether the measurement holds up.
Sample size and duration come out of assumptions you declare, and the skill asks for the approximations to be stated rather than buried. One decision surface changes at a time unless the design explicitly estimates interactions. Exclusions, quality checks, the analysis plan and decision thresholds get pre-registered before launch, not after.
On readout it verifies assignment integrity and data completeness first, then estimates the effect with its uncertainty. You get versioned JSON and a decision in ordinary words.
The skill also names the habits that ruin these tests: peeking repeatedly and stopping when the number looks good, and calling underpowered noise a winner. Generalising past the tested population is the third. Worth checking your process against all three before the budget goes out.
What it helps with
- plan an ad A/B test with a stated hypothesis and randomization unit
- size the sample and duration from declared assumptions
- pre-register stopping rules and exclusions before launch
- audit a readout for assignment integrity and data completeness
- separate an underpowered result from a real lift
Example prompts for the agent
- Design an A/B test for two creatives: hypothesis, treatment and control, primary metric, sample size and duration from my assumptions.
- Check this test design for overlapping experiments and conversion lag, then return the setup as JSON.
- Read out the test: assignment integrity and data completeness first, then the effect with uncertainty.
How to install ads-test
1Skills CLI (Vercel)
npx skills add AgriciDaniel/claude-ads --skill ads-test -a claude-code2Manually: put the files in place
~/.claude/skills/ads-test/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/ads-test/
1Skills CLI (Vercel)
npx skills add AgriciDaniel/claude-ads --skill ads-test -a codex2Manually: put the files in place
.agents/skills/ads-test/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add AgriciDaniel/claude-ads --skill ads-test -a cursor2Manually: put the files in place
.cursor/skills/ads-test/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install AgriciDaniel/claude-ads ads-test2Manually: put the files in place
.github/skills/ads-test/SKILL.md1Skills CLI (Vercel)
npx skills add AgriciDaniel/claude-ads --skill ads-test -a gemini-cli2Platform CLI
gemini skills install https://github.com/AgriciDaniel/claude-ads --path skills/ads-testManually: put the files in place
.devin/skills/ads-test/SKILL.mdThe legacy .windsurf/skills/ path still works
Skills CLI (Vercel)
npx skills add AgriciDaniel/claude-ads --skill ads-test -a opencodeClaude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
From the same set20
The whole set →Similar
Questions about ads-test
How do I install ads-test in Claude Code?
Run npx skills add AgriciDaniel/claude-ads --skill ads-test -a claude-code. The skill lands in ~/.claude/skills/ads-test/SKILL.md.
What does ads-test do?
It runs a paid-ad experiment step by step: hypothesis, randomization unit, guardrails, sample-size and duration maths, pre-launch checks, then a readout with effect and uncertainty. Output is versioned JSON plus a decision in plain language.
What does it check during a readout?
Assignment integrity and data completeness come before any effect estimate. If the test is underpowered, the result is called noise instead of a winner.
Source text
SKILL.md · MIT
--- name: ads-test description: "Design and evaluate paid-ad experiments with hypotheses, randomization units, sample-size and duration assumptions, guardrails, platform experiment tools, analysis, and decision rules. Use for A/B test, split test, experiment design, hypothesis, statistical significance, sample size, test duration, or experiment readout." --- # Paid Media Experiment 1. State the decision, causal hypothesis, treatment, control, randomization unit, population, primary metric, guardrails, minimum effect, and stopping rule. 2. Check platform constraints, overlapping experiments, conversion lag, seasonality, interference, and measurement quality. 3. Calculate sample and duration from declared assumptions; disclose approximations. 4. Change one decision surface unless the design explicitly estimates interactions. 5. Pre-register exclusions, quality checks, analysis, and decision thresholds. 6. For readout, verify assignment integrity and data completeness before estimating effect and uncertainty. 7. Return setup or readout in versioned JSON with a plain-language decision. Do not repeatedly peek and stop on a favorable result, call underpowered noise a winner, or generalize beyond the tested population.
© AgriciDaniel · MIT · AgriciDaniel/claude-ads