Skip to content
SkillModel-led167 tokens in use

ads-test

Designs paid ad experiments with hypotheses, sample size, duration, guardrails and stopping rules, then reads results out with assignment and data checks.

11,191 stars on the repository MIT LicenseAgriciDaniel/claude-ads · skills/ads-test
/plugin marketplace add agricidaniel/claude-ads /plugin install claude-ads@ai-marketing-hub-claude-ads
View source

Install command as written in the repository's README

Skill label

measured from its files

What installing ads-test costs

63tokens idle
167tokens in use

Idle is what sits in context while the skill is installed. In use is its instructions, loaded when it runs. Counts are approximate, made offline with cl100k_base.

Model dependence Model-led

Script-ledMixedModel-led

Mostly instructions, so how well it works depends on the model that follows them.

This describes the skill's structure, how much of it is scripts. It does not say how well it works.

Touches

Nothing detected: no shell, network, file writes, git or credentials in its files.

Declared tags come from the skill's frontmatter; detected ones from its scripts, its code blocks and the commands it names in inline code. This is detected, not audited.

How it scores

Quality

How well the skill is written, counted from its own files: its frontmatter, its steps, examples, stated output and constraints, its size and whether every file it ships is one it names.How we score it →
Fair
  • Frontmatter8/15
  • Structure11/25
  • Examples0/15
  • Output12/15
  • Size15/15
  • Resources10/15

Security

Starts at 100 and loses points for each thing in the skill's files that a reader should look at first.How we score it →
Excellent

No findings.

Repository health

Pre-computed score based on development activity, maintenance, community, maturity, and trend momentum.How we score it →
Good
  • Development activity84
  • Maintenance100
  • Community56
  • Maturity32
  • Momentum40

ads-test covers both ends of a paid media experiment. For setup it states the decision, causal hypothesis, treatment, control, randomization unit, population, primary metric, guardrails, minimum effect and stopping rule, and checks platform constraints, overlapping experiments, conversion lag, seasonality and interference.

For a readout it verifies assignment integrity and data completeness before estimating the effect and its uncertainty. Either way the result comes back as versioned JSON with a decision written in everyday language.

What You Get

  • A test design with hypothesis, treatment, control, randomization unit and primary metric
  • Sample size and duration estimates with their assumptions stated
  • Guardrails, minimum effect and a stopping rule
  • A readout with assignment integrity and data completeness checks
  • Versioned JSON with a decision in everyday language

Common Use Cases

  • Planning an A/B test of ad creative or audiences
  • Deciding how long a split test must run
  • Reading out a finished platform experiment
  • Checking whether overlapping experiments or conversion lag undermine a result

What Makes It Unique

The instructions rule out common experiment mistakes: peeking repeatedly and stopping on a favorable result, calling underpowered noise a winner, and generalizing beyond the tested population. Exclusions and decision thresholds are fixed before the test, and approximations in the sample size math are disclosed.

Under The Hood

The skill is a short set of written instructions with no scripts of its own. The agent works from the test setup or results you provide and does the sample size calculation itself from your assumptions. It returns versioned JSON and does not call web services or run commands. It does not change any ad account or start a platform experiment.

Join founders building with open source

Opinionated takes, migration guides, cost-saving tips, and insights from the open source ecosystem.

Subscribe on SubstackJoin 750+ subscribers