Content
35%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
This SKILL.md functions almost entirely as a table of contents with no substantive content in the body itself. It lacks any actionable guidance, concrete examples, or executable steps—everything is deferred to 11 sub-skill files that aren't provided. While the sequential structure hints at a reasonable workflow, the absence of inline summaries, validation criteria, or any concrete instructions makes the skill body insufficient as a standalone reference.
Suggestions
Add a concise end-to-end workflow summary in the body (e.g., numbered steps with the key action and gate criteria for each phase) so the skill is useful even without opening sub-files.
Include at least one concrete, actionable example—such as a sample hypothesis statement that passes the quality checklist, or a sample size calculation command/formula.
Add brief 1-sentence descriptions next to each sub-skill link explaining what it covers and when to use it, rather than bare links.
Define the 'hard gate' criteria inline (e.g., what specifically must be true to pass the Hypothesis Lock or Execution Readiness Gate) so Claude can act on them without navigating to sub-files.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Reasonably brief at the top level, but includes some unnecessary framing ('Ensure every A/B test is valid, rigorous, and safe before a single line of code is written') and emoji decoration that adds no value. The bullet points under Purpose & Scope are vague rather than instructive. | 2 / 3 |
Actionability | The SKILL.md contains no concrete guidance, no executable steps, no code, no commands, and no examples. It is entirely a table of contents pointing to sub-skill files, with no actionable content in the body itself. | 1 / 3 |
Workflow Clarity | The numbered sub-skill references imply a sequence (hypothesis → assumptions → test type → metrics → sample size → execution gate → during test → analysis → interpretation → record), and the 'hard gate' labels suggest validation checkpoints. However, no explicit workflow steps, validation criteria, or feedback loops are described in the body—everything is deferred to sub-files. | 2 / 3 |
Progressive Disclosure | The skill does attempt progressive disclosure by linking to 11 sub-skill files, which is a reasonable structure. However, the links have no descriptions or summaries explaining what each module covers, making navigation harder. Additionally, no bundle files were provided, so we cannot verify the references resolve, and the sheer number of sub-files (11) without any inline summary content makes the top-level file nearly useless on its own. | 2 / 3 |
Total | 7 / 12 Passed |