Content
65%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.
The body is admirably lean and dense with package-specific knowledge Claude would not already have, and its guardrails are unusually substantive. Its weaknesses are the non-executable Core Pattern skeleton, the absence of any ordered workflow or validation checkpoints for budget-consuming optimizer runs, and references to bundle files that are not actually present.
Suggestions
Replace the Core Pattern skeleton with one fully executable example that shows how reflection_client, request, and evaluator are actually constructed (e.g., drawn from the package examples), so the snippet is copy-paste ready.
Add a short numbered workflow for a typical GEPA run (set metric budget → bootstrap demos if needed → run optimize → inspect artifact metadata for baselineScore vs bestScore) with an explicit verification step on the resulting Pareto front.
Either ship the referenced files (API.md, axir-api.json, axir-capabilities.json, examples/) in the bundle or inline the minimal needed content, so the body's pointers resolve to real resources.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is lean and assumes competence: bullet-style Package Facts, a four-line Core Pattern, and guardrails consisting entirely of non-obvious package semantics (baseline score placement, metric-budget accounting, demo freshness) that Claude could not know. Only trivial redundancy exists ("Language: Python" restating the description), which does not rise to the over-explanation of a 4. | 5 / 5 |
Actionability | The Core Pattern is a skeleton with undefined placeholders — `AxGEPA(reflection_client)` and `engine.optimize(request, evaluator)` never show how to construct the reflection client or evaluator — so it functions as an API shape sketch rather than executable code. There is some concrete guidance (the API surface list and 'Start from package examples'), matching the pseudocode-with-missing-details anchor rather than the mostly-executable anchor at 4. | 3 / 5 |
Workflow Clarity | Content is organized by topic (When To Use, Package Facts, Core Pattern, Guardrails) with no sequenced process and no validation checkpoints, despite covering budget-consuming optimizer runs where a checkpoint discipline would help. It is coherent with clear usage triggers — above the rough-sequence-with-many-gaps anchor at 2 — but the sequence-present-but-checkpoints-missing anchor at 3 is the best fit given guidance exists without an ordered flow. | 3 / 5 |
Progressive Disclosure | Sections are well organized for a sub-50-line skill, but the referenced resources — "API.md and axir-api.json", "axir-capabilities.json", "examples/" — do not exist in the bundle, so the pointers cannot actually be followed. This lands on the some-structure-with-imperfectly-usable-references anchor at 3 rather than 4, since navigation to deeper detail is asserted but unverifiable. | 3 / 5 |
Total | 14 / 20 Passed |