Content
86%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 well-engineered skill body: lean, expert-level prose, executable scaffold and hook examples, and excellent reference structure with verified one-level-deep bundle files. The main gaps are the missing emdash-plugin.jsonc manifest example for the capability declarations it emphasizes and an implicit rather than explicit step sequence.
Suggestions
Add a minimal emdash-plugin.jsonc example showing one or two capability declarations, since 'Declare access' is a core step but currently has no concrete template.
Make the build-then-test sequence explicit (e.g., a short numbered overview of scaffold → declare → implement → test) so the workflow order does not rely on section arrangement alone.
Include a one-line example of declaring a route's methods and body modes, which the 'Routes, HTTP, and MCP' section requires but never concretizes.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense declarative prose with zero padding ('The manifest is the identity and trust contract', 'Sandboxed hooks enter the same priority, dependency, timeout, error-policy, enablement, exclusive-provider, and capability pipeline as trusted hooks') and explains no concepts Claude already knows — every token earns its place. | 5 / 5 |
Actionability | Provides copy-paste scaffold commands and a complete executable src/plugin.ts example, plus exact capability strings and named test-host functions. Not 5: it instructs 'Declare every host API in emdash-plugin.jsonc' without showing a manifest example, and route/MCP declaration syntax is delegated to references. | 4 / 5 |
Workflow Clarity | Section order follows the development lifecycle (choose format → scaffold → declare access → routes/hooks/UI → runner parity → test) and the testing section explicitly separates the two host types with when-to-use guidance and disposal requirements. Not 5: the sequence is implicit in section order rather than an explicit stepped workflow with validation checkpoints. | 4 / 5 |
Progressive Disclosure | SKILL.md is a clear overview with 12 focused, well-signaled one-level-deep references — all verified to exist — signposted both inline per topic ('Read [Hooks](./references/hooks.md)') and in a consolidated References section; cross-links inside references are sibling-level, not nested chains. | 5 / 5 |
Total | 18 / 20 Passed |