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 strong, well-structured reference: nearly every line is an executable command or a non-obvious gotcha, workflows are gated by bootstatus readiness with error-recovery escalation and a review checklist, and detail is correctly offloaded to a real, clearly-signaled one-level-deep bundle file. The only deductions are minor: slight repetition of the bootstatus guidance and no explicit post-install/launch verification step.
Suggestions
State the bootstatus-as-readiness-gate guidance once (in Device Lifecycle or App Install) and reference it from the other section instead of repeating it.
Add a one-line verification step after install/launch (e.g., confirm the app process is running via `simctl spawn booted launchctl list | grep <bundle-id>` or check `get_app_container` succeeds) to close the workflow's validation loop.
Merge the two adjacent sentences in Location Simulation that both explain the set/run/start command boundary into a single tight explanation.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is overwhelmingly executable commands with terse, non-obvious annotations (booted non-determinism, SIGKILL losing video, alpha mask screenshot-only), assuming Claude's competence rather than explaining basics. It sits at the score-4 anchor 'efficient; minor instances of over-explanation that could be trimmed' — e.g., the bootstatus-as-readiness-gate guidance appears twice (lines 93-94 and 107-108) and the location section explains the set/run/start split across two overlapping sentences — rather than the score-5 'every token earns its place'. | 4 / 5 |
Actionability | Fully executable, copy-paste-ready bash throughout: complete xcodebuild/install/launch invocations, a concrete JSON push payload with both file and stdin delivery, jq pipelines for UDID lookup, Swift targetEnvironment examples, and WRONG-vs-CORRECT command pairs. Specific examples cover the common cases, matching the score-5 anchor rather than the score-4 anchor's 'minor gaps'. | 5 / 5 |
Workflow Clarity | Sequences are clear with most checkpoints present: the install flow gates on `bootstatus <UDID> -b`, the stuck-boot flow shows the exact error string with a shutdown→erase→retry→CoreSimulator-reset escalation (a feedback loop), CI cleanup uses `trap cleanup EXIT`, and a Review Checklist closes the skill. It stays at 4 rather than 5 because there is no explicit verification that install/launch actually succeeded (e.g., checking the process or app state) — a minor validation gap; it does not drop to 3 since destructive commands (erase/delete) are labeled with their effects and embedded in recovery/teardown sequences. | 4 / 5 |
Progressive Disclosure | The body is an overview of common cases while exhaustive material (complete command tables, jq parsing, full privacy service list, Logger filter setup, boot recovery commands) is split into one-level-deep references/simctl-commands.md, which exists and contains the linked `#logger-and-filtered-streaming` anchor. References are clearly signaled up front ('Load [the simctl command reference]... when you need complete command tables') and re-linked contextually, matching the score-5 'clear overview with well-signaled one-level-deep references'. | 5 / 5 |
Total | 18 / 20 Passed |