Content
78%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.
An excellent rules/reference skill for code generation: executable examples for every feature, explicit defaults and anti-patterns, and no filler concept explanations. Its main weaknesses are monolithic single-file structure with inlined type/API reference material, one padded statistics passage, and rule repetition that could be deduplicated or moved to a reference file.
Suggestions
Split the full options/type reference (the 'Types' listings, AxAgentCatalogSkill/AxAgentContextMap types) and cross-language persistence notes into a references/ file, keeping SKILL.md to defaults, key examples, and rules.
Drop the A/B gate statistics sentence from the relevanceRanking section; keep only the behavioral contract (default ON, opt out with false, non-authoritative hint).
Deduplicate the repeated rules — 'return void', do-not-loop, and child-agent non-inheritance each appear in both 'Use These Defaults' and 'Do Not Generate'; state each once and cross-reference.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense with API-specific contracts ('Catalog content is NOT preloaded into the prompt; entries load only when recalled') and avoids explaining known concepts, but includes trimmable padding — the relevanceRanking A/B statistics sentence ('substance-judged, 49 runs per variant per model: small-model first-lookup precision 24%→90%...') is irrelevant to code generation, and the void-return/no-loop/child-agent rules repeat across 'Use These Defaults', actor usage rules, and 'Do Not Generate'. | 4 / 5 |
Actionability | Every feature ships copy-paste-ready TypeScript: AxAgentContextMap construction with onUpdate persistence, an onMemoriesSearch implementation with alreadyLoaded dedup, catalog arrays, forward-time skill/memory preloading, a carried-memories pattern, and observer wiring, plus the supporting type definitions — fully executable and covering the common cases. | 5 / 5 |
Workflow Clarity | 'Use These Defaults' provides a clear decision entry point (static catalog vs. host callback vs. contextMap), each section follows an Enabling → Actor usage → Rules sequence, and 'Do Not Generate' acts as an explicit anti-pattern checklist. It stops short of a 5 because there is no explicit ordered decision flow for combining features in one agent. | 4 / 5 |
Progressive Disclosure | The body is a well-sectioned but monolithic ~440-line single file: the full options/type reference, relevanceRanking details, and cross-language persistence notes are inlined rather than split into reference files. The only external pointers are clearly signaled one-level-deep GitHub example URLs, and no bundle (references/, scripts/, assets/) exists to offload the API reference material. | 3 / 5 |
Total | 16 / 20 Passed |