Content
68%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 terse, well-organized body that respects the token budget and sets useful guardrails. Its main weaknesses are placeholder-laden code that is not copy-paste executable and the absence of explicit validation/rollback checkpoints for the evolve workflow it promotes.
Suggestions
Make the Core Pattern executable by showing how 'llm', 'examples', and 'metric_fn' are constructed (e.g. via a no-key example) or by pointing to one specific runnable example file path.
Add a short numbered workflow for evolve/refine/render with an explicit validation checkpoint (e.g. 'only accept evolved rules when verification passes; otherwise rely on exact rollback').
Give each optimizer name in 'Relevant API Surface' a one-line role, or move the list into a pointer to a specific API.md section.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is lean and fact-dense ('Language: Python.', 'Package: `axllm`.') with no explanations of concepts Claude already knows and no padding. Every section earns its tokens. | 5 / 5 |
Actionability | The Core Pattern shows real call shapes but relies on undefined placeholders ('llm', 'examples', 'metric_fn'), making it pseudocode-like rather than executable. It defers executable detail to external package examples instead of including it, matching the 3 anchor better than the 'minor gaps' 4 anchor. | 3 / 5 |
Workflow Clarity | The Core Pattern implies a rough sequence (ax -> playbook -> evolve) and 'When To Use' enumerates tasks, but no explicit checkpoints or validation/rollback guidance is given for the risky evolve operation. Sequence present with implicit checkpoints, matching the 3 anchor. | 3 / 5 |
Progressive Disclosure | The ~44-line body is cleanly sectioned and the skill ships no bundle files, but detail pointers ('API.md', 'axir-api.json', 'examples/') name package files rather than clearly-signaled one-level-deep skill references, and the 'Relevant API Surface' is a bare inline name list. Good structure with minor organization gaps — the 4 anchor. | 4 / 5 |
Total | 15 / 20 Passed |