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 tight, highly actionable skill body that assumes Claude's competence and provides copy-paste commands plus explicit error-recovery guidance. The only minor gap is that the core workflow's validation step is implicit rather than an explicit checkpoint.
Suggestions
Make the happy-path validation explicit, e.g. after 'agent-device open ...', add a step like 'Verify the app foregrounded via the interactive snapshot before proceeding.' to lift workflow clarity to a 5.
Consider a short numbered sequence for the normal-task flow (open → verify snapshot → act) so the ordering and checkpoint stand out at a glance.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Lean and efficient throughout — it never explains what an iOS Simulator, React Native, or the CLI is; every line is operational guidance (install guardrail, do-not-probe list, open command, error handling) that earns its place. | 5 / 5 |
Actionability | Fully executable, copy-paste-ready commands — 'npm install -g agent-device@latest', 'agent-device open <app-or-bundle-id> --platform ios --foreground', and a block of help topics with inline comments — cover the common cases. | 5 / 5 |
Workflow Clarity | Clear sequence (require install → open with --platform ios → follow snapshot/error hints → stop-and-ask if unavailable → read help only when specialized) with explicit error-recovery feedback, but the happy-path validation is implicit ('follow the interactive snapshot') rather than an explicit verify checkpoint, leaving a minor gap. | 4 / 5 |
Progressive Disclosure | A self-contained 28-line skill under the 50-line threshold with no need for external reference files; content is logically grouped (install, normal driving, error handling, specialized help) and the help-topic pointers are one-level navigation into the tool's own help system. | 5 / 5 |
Total | 19 / 20 Passed |