CtrlK
BlogDocsLog inGet started
Tessl Logo

testland/feature-flag-experiment-validator

Validates the statistical significance of an A/B / feature-flag experiment result - computes per-metric effect size + p-value (chi-square for proportions, Welch's t-test for continuous metrics), applies a multiple-comparison correction (Bonferroni / Benjamini-Hochberg) when N>1 metric, surfaces practical-vs-statistical-significance distinction, and emits a ship/don't-ship verdict per metric. Use when an experiment has finished and someone is about to ship the winning variant off a dashboard readout, when a result rests on a small sample, or when more than one metric was compared - the rigorous version of "the variant looks better in the dashboard."

75

Quality

94%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

SecuritybySnyk

Passed

No findings from the security scan

Overview
Quality
Evals
Security
Files

Quality

Content

85%

Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.

A dense, well-structured skill body: executable code throughout, a clearly gated seven-step workflow, and clean progressive disclosure into two real reference files. The only weakness is a brief Overview that restates A/B-testing fundamentals Claude already knows.

Suggestions

Trim or relocate the Overview's explanation of A/B-test variance/sample-size basics; keep only the citation anchors and assume Claude's familiarity to push conciseness to a 3.

Consider moving the false-positive probability aside (1 - 0.95^10 = 40%) into the anti-patterns reference if it is motivational rather than load-bearing for the correction step.

DimensionReasoningScore

Conciseness

The Overview paragraph explains A/B-testing basics Claude already knows ('A/B tests are sensitive to variance; they require a large sample size…'), which the rubric penalizes; the rest is lean code and direct steps, so it sits at 'mostly efficient but includes some unnecessary explanation' rather than fully lean.

2 / 3

Actionability

Provides fully executable, copy-paste-ready code (chi2_contingency, ttest_ind, mannwhitneyu, multipletests, NormalIndPower) plus concrete YAML input and MDE formats — matching the score-3 'fully executable code' anchor.

3 / 3

Workflow Clarity

Steps 1–7 are clearly sequenced with explicit validation gates (power-analysis gate in Step 4, 'both statistical AND practical significance' in Step 5, peeking→sequential testing in Step 7) and feedback loops (inconclusive→'re-run at higher N'), matching the score-3 anchor.

3 / 3

Progressive Disclosure

The body is an overview that splits detail into two real, one-level-deep, well-signaled references (references/output-example.md, references/anti-patterns.md, both verified present), matching the score-3 'clear overview with well-signaled one-level-deep references' anchor.

3 / 3

Total

11

/

12

Passed

Description

100%

Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.

A strong, third-person description that names concrete capabilities and pairs them with natural, well-scoped 'Use when' triggers. It clearly communicates both what the skill does and when to invoke it, with low conflict risk.

DimensionReasoningScore

Specificity

Lists multiple concrete actions — 'computes per-metric effect size + p-value (chi-square…Welch's t-test)', 'applies a multiple-comparison correction (Bonferroni / Benjamini-Hochberg)', 'emits a ship/don't-ship verdict per metric' — matching the score-3 anchor rather than the partial score-2.

3 / 3

Completeness

Explicitly answers both what ('Validates the statistical significance…emits a ship/don't-ship verdict') and when ('Use when an experiment has finished…'), satisfying the score-3 anchor with explicit triggers.

3 / 3

Trigger Term Quality

The 'Use when…' clause surfaces natural terms a user would actually say — 'ship the winning variant off a dashboard readout', 'small sample', 'more than one metric', 'A/B / feature-flag experiment', 'the variant looks better in the dashboard' — giving good coverage despite the statistical jargon in the capability half.

3 / 3

Distinctiveness Conflict Risk

Occupies a clear niche (statistical validation of finished experiments) with distinct triggers unlikely to fire for other skills; not generic like the score-1 'helps with code and documents' example.

3 / 3

Total

12

/

12

Passed

Validation

100%

Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.

Validation16 / 16 Passed

Validation for skill structure

No warnings or errors.

Reviewed

Table of Contents