Content
93%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 disciplined overview skill: lean and dense with contracts and boundaries, executable where it counts (scaffold commands, complete hook example, exact test-host functions), with detail correctly delegated to a well-organized one-level-deep reference bundle. The only weakness is that the end-to-end plugin-building workflow is conveyed by section ordering and a testing section rather than explicit numbered steps with a validation-recovery loop.
Suggestions
Make the build sequence explicit — e.g., a short numbered workflow (choose format → scaffold → declare access → implement routes/hooks/UI → test) at the top, instead of relying on section order to imply it.
Add one feedback-loop line to the testing section (e.g., "On test failure, fix and re-run before extending coverage") to close the validation-retry gap for destructive capabilities such as content:publish and media:write.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The ~130-line body carries only non-obvious, decision-critical information ("Sandboxed runtime code can use Web APIs but not Node.js built-ins", "Patch does not imply read", the 8 MiB body bound) and assumes Claude's competence with no padding about general concepts. The only redundancy — five references linked both inline and again in the References list — is navigation rather than over-explanation, so it stays at anchor 5 rather than 4. | 5 / 5 |
Actionability | Provides copy-paste-ready scaffold commands ("pnpm dlx @emdash-cms/plugin-cli init my-plugin") and a complete, typed `src/plugin.ts` example, plus exact identifiers throughout ("createPluginTestHost()", "host.http.respond()", "ctx.storage.<collection>", import paths "emdash/plugin"). These fully cover the common cases; the one omission (no inline `emdash-plugin.jsonc` manifest snippet) is a shape the CLI generates, so it does not fall to anchor 4's "minor gaps". | 5 / 5 |
Workflow Clarity | Sections follow the natural build order — Choose a format → Scaffold → Declare access → Routes/Hooks/Admin UI → Runner parity → Test the production boundary — with real validation checkpoints (the two test-host tiers, "Dispose every host after use", `pnpm run test` in the scaffold). It is anchor 4 rather than 5 because the sequence is implied by section order rather than explicit numbered steps, and there is no fix-and-retry feedback loop or checklist; it stays above anchor 3 because validation is concrete and present, not missing. | 4 / 5 |
Progressive Disclosure | The overview holds decisions while ~2,300 lines of per-API detail are split across 12 real reference files, each well-signaled inline from the section that needs it ("Read [API routes and MCP tools](./references/api-routes.md)") and in a consolidated list. References are one level deep with only sibling-level cross-links (e.g., admin-ui.md → block-kit.md), matching anchor 5; the inline 28-row capability table is summary-level decision content, not misplaced detail. | 5 / 5 |
Total | 19 / 20 Passed |