Content
71%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 highly actionable, well-organized reference packed with executable, SDK-specific guidance and validation tooling. Its weaknesses are verbosity from duplicated sections and an inlined API reference that should live in bundled reference files, plus the absence of a single unified build/validate/test sequence.
Suggestions
Move the World SDK method signatures, event-types table, and dashboard URL-format reference into a references/ file (e.g., references/world-sdk.md, references/cli.md) and keep only the key signatures and pointers in SKILL.md, cutting the file to well under half its current size.
Merge the "Error handling" and "Error handling patterns" sections into one, since both demonstrate FatalError/RetryableError with near-identical code.
Collapse the four near-identical namespaced-stream writer functions (logDebug/logInfo/emitAgentThought/emitAgentResult) into one parameterized helper example, stating the namespace naming convention in a sentence instead.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense with SDK-specific, non-obvious knowledge and avoids explaining basics Claude already knows, but it includes clear padding: the FatalError/RetryableError guidance appears in two separate sections ("Error handling" and "Error handling patterns") with near-duplicate code, and the namespaced-streams example repeats the identical try/finally writer pattern across four functions (logDebug, logInfo, emitAgentThought, emitAgentResult). It could be tightened to a 4-5 by deduplicating these, but as written it matches "mostly efficient with some unnecessary content". | 3 / 5 |
Actionability | Nearly every section gives copy-paste-ready, complete TypeScript or shell commands: directives, imports, step/workflow patterns, hook creation and resumption, serde class protocol, chunking/batching patterns, vitest integration config with waitForHook/waitForSleep, CLI commands with flags (--json, --url, --backend), and World SDK method signatures. The examples cover the common cases end to end. | 5 / 5 |
Workflow Clarity | There is a clear docs-first procedure (numbered glob/grep steps against node_modules/workflow/docs/), explicit validation tooling for serde compliance (workflow transform --check-serde, workflow validate --json, build-time warnings), and testing guidance with a worked integration-test flow. It falls short of a 5 because there is no unified end-to-end sequence (write -> validate -> test) tying the sections together; validation checkpoints exist per-topic rather than as an explicit feedback loop. | 4 / 5 |
Progressive Disclosure | The SKILL.md is an 845-line monolith with no bundled references/ directory, and roughly 150+ lines of it are API-reference material inlined inline (World SDK method signatures, the event-types table, streams methods, dashboard URL formats) that clearly belongs in a separate reference file — closely matching the anchor-3 example. Pointers to node_modules/**/docs/ are clearly signaled and only one level deep, which keeps it above a 2, but the skill bundle itself makes no use of progressive disclosure. | 3 / 5 |
Total | 15 / 20 Passed |