Content
88%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, disciplined body: executable commands throughout, an explicit validated workflow with failure conditions and feedback loops, and no padding or basic-concept explanations. The main gaps are minor — redundant overlap between Common Mistakes and the Review Checklist, and a largely monolithic layout that underuses reference files despite a well-organized section structure.
Suggestions
Consolidate 'Common Mistakes' and 'Review Checklist' into one section (or trim the checklist to items not already covered as mistakes) to remove redundancy and save tokens.
Move the 'Interpret the Report' semantics detail into a reference file (e.g., references/interpretation.md) with a one-line pointer, keeping SKILL.md as a lean overview with quick-start workflow only.
Add the full device-capture command alongside the Simulator one (or a short flag table) so the device path is as copy-paste ready as the Simulator path.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body assumes Claude's competence — no basic explanations of profiling or symbolication — and every command block is bare and purposeful. Minor trimming is possible: 'Common Mistakes' and 'Review Checklist' overlap substantially (e.g., unbounded sessions, mismatched dSYMs, single-run claims appear in both), and framing sentences like "Treat capture conditions and symbolication as part of the evidence, not as setup trivia" could be tightened. Not score 5 due to this minor redundancy; not 3 since there is no padded or already-known explanatory content. | 4 / 5 |
Actionability | Commands are fully executable and copy-paste ready: 'brew install emergetools/homebrew-tap/ettrace', 'python3 scripts/collect_dsyms.py --app ... --search-root ... --output ... --pretty', 'ettrace --simulator --dsyms /tmp/myapp-ettrace/dsyms', and 'python3 scripts/analyze_ettrace.py ... --top 25 --pretty'. Both referenced scripts exist in the bundle. Concrete failure conditions ("The second mkdir must fail", "Missing or ambiguous UUID matches stop the run by default") make the guidance directly actionable; only the device-capture variant is given as a delta ('omit --simulator') rather than a full command, which is a reasonable, explicitly justified simplification. | 5 / 5 |
Workflow Clarity | Six numbered steps are clearly sequenced (triage → wire runner with fixed versions → build + collect UUID-matched dSYMs → capture in a clean directory → validate/summarize → change one cause and repeat) with explicit validation checkpoints: confirm the 'Starting ETTrace' log, dSYM UUID matching that stops on failure, the must-fail mkdir guard, and "Stop and repair symbolication when important app frames are <unknown>". Feedback loops (validate → repair → re-capture, and repeat-at-least-twice verification) plus a review checklist match the anchor-5 pattern. | 5 / 5 |
Progressive Disclosure | Structure is good: a Contents nav, clearly delimited sections, executable logic correctly externalized into the two real bundle scripts (both referenced paths exist), and a References section of one-level-deep external links. It falls short of anchor 5 because the ~190-line body is essentially monolithic — detail like the 'Interpret the Report' semantics and the 13-item checklist could live in a reference file with clear in-body pointers, and no bundle reference files are used for that split. It is clearly above anchor 3, where structure would be unclear or references buried. | 4 / 5 |
Total | 18 / 20 Passed |