Content
82%Weight 40%Scale 1-5Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
A strong, highly actionable codegen skill: nearly every line is library-specific knowledge Claude cannot infer, anchored by two complete canonical examples and concrete troubleshooting. The main weaknesses are minor redundancy between the defaults list and the examples, and the non-portable absolute example paths in place of a proper reference structure.
Suggestions
Replace the absolute /Users/vr/src/ax/... paths in 'Good Example Targets' with portable references (repo-relative paths or bundled example files under references/), since no other user can resolve those paths.
Consolidate the option explanations in 'Use These Defaults' and 'Useful Options' — they overlap with the inline options in the two canonical examples — to trim roughly a dozen lines without losing information.
Consider moving the full AxGEPA constructor options reference and the example-target listing into a references/ file, keeping SKILL.md as a lean overview with well-signaled one-level-deep links.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense with non-inferable library specifics (baseline semantics, bootstrap trace rules, isExpensive teacher handling) rather than concepts Claude already knows, but there is minor redundancy — the 'Use These Defaults' list overlaps the canonical examples' option sets, and the three long baseline-semantics bullets in Critical Rules could be tightened. | 4 / 5 |
Actionability | Two complete, copy-paste-ready TypeScript examples (scalar and Pareto) include imports, model setup, metrics, full option objects, and result application, and the Troubleshooting section gives concrete fixes like "increase it until the initial validation pass fits" and `teacherOptions: { useExpensiveModel: 'yes' }`. | 5 / 5 |
Workflow Clarity | The Metric Selection section and 'Rule of thumb' clearly route between `optimize(...)`, plain evaluators, and `agent.optimize(...)`, and Troubleshooting provides error-recovery feedback loops; however, the guidance is presented as rules rather than an explicit sequenced workflow with checkpoints, leaving minor gaps. | 4 / 5 |
Progressive Disclosure | A single-file skill with no bundle, organized into clear, logically ordered sections (defaults, critical rules, metric selection, patterns, budgeting, troubleshooting). At ~270 lines, the full options reference and six example targets could live in a separate reference file, and 'Good Example Targets' cites non-portable absolute paths (`/Users/vr/src/ax/...`) that other users cannot access. | 4 / 5 |
Total | 17 / 20 Passed |