Content
82%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 tight, highly actionable skill body: executable quick start, explicit when-to-use/when-not-to-use guidance, and high-value anti-patterns with WHY explanations. The main weaknesses are minor redundancy in test instructions, implicit rather than sequenced workflow steps, and navigation gaps — three dead eval links and most bundle reference files being undiscoverable from SKILL.md.
Suggestions
Fix or remove the three dangling eval-scenario links: the body references evals/scenario-0..2/task.md but no evals/ directory exists in the bundle.
Add a short references index in the body (e.g., publishing → references/publishing.md, testing → references/testing.md, UI feedback → references/ui-feedback.md) so the rest of the bundle is discoverable from SKILL.md.
Consolidate the duplicate test instructions — merge 'Then test: opencode run hi' and the 'Verify plugin is loaded' command into a single verification step in Quick Start.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The ~85-line body is lean with no concept re-teaching and executable code throughout, but has minor trims available: duplicate test commands ('Then test: `opencode run hi`' vs 'Verify plugin is loaded: run `bun run opencode run ...`') and the unclear 'Optional — document plugin purpose in a companion agent file' snippet. | 4 / 5 |
Actionability | Fully executable guidance: copy-paste-ready plugin factory code, `opencode.json` registration config, concrete test commands, and BAD/GOOD anti-pattern pairs with runtime consequences explained ('NEVER call `client.registerTool()` — it does not exist.'). | 5 / 5 |
Workflow Clarity | Quick Start establishes a create → register → test sequence with an explicit verification checkpoint ('Verify plugin is loaded: run `bun run opencode run "list your tools"` and confirm the plugin output appears') and anti-patterns as recovery guidance; falls short of 5 because steps are prose-implicit rather than a sequenced workflow with a fix-and-retry loop. | 4 / 5 |
Progressive Disclosure | The body is a genuine overview with a clearly signaled one-level-deep pointer ('See references/hook-patterns.md for complete anti-pattern list + full hook/event/tool/UI/testing/publishing reference') that exists and delivers. Docked from 5: the three eval-scenario links point to nonexistent `evals/scenario-*/task.md` paths, and 9 of the 11 bundle reference files (publishing.md, testing.md, ui-feedback.md, events.md, etc.) are not navigable from the body. | 4 / 5 |
Total | 17 / 20 Passed |