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 well-structured tool-orchestration skill: lean tables, concrete commands, explicit recovery rules, and properly split one-level references. The main defects are the duplicated tool table / repeated delegation note and a Read-vs-grep contradiction inside the console-log workflow.
Suggestions
Resolve the contradiction in section 5: workflow step 2 ('Search the file using Grep or Read') conflicts with the later rule ('Never Read the log file directly') — pick one (grep/shell only) and state it once.
Deduplicate the repeated 'delegate to an Explore subagent' advice in section 5, and consider dropping or trimming the Quick Reference table, which repeats ~8 tool rows already covered by the Tool Overview.
Add one concrete example invocation (e.g., a sample debugger-inspect-element call with logical-pixel x/y coordinates) to make the most-used inspection flow copy-paste ready.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Efficient and dense — tables, a copy-paste adb command, and no filler explaining React Native or CDP basics. Not a 5: the Quick Reference table duplicates ~8 rows of the Tool Overview, and the 'delegate to an Explore subagent' note appears twice in section 5, so there is trimming to do without losing information. | 4 / 5 |
Actionability | Fully concrete and executable: 'adb -s <serial> reverse tcp:8081 tcp:8081', specific numeric limits ('start at 50, then grow', '-m 50', 'tail -N'), an exact flat-log field spec, and unambiguous tool selection guidance cover the common cases. | 5 / 5 |
Workflow Clarity | Clear sequencing with a Golden Rules checklist, an upfront validation step ('Verify via debugger-status'), and feedback loops (status -> restart-app -> retry status). Not a 5: the log workflow is internally contradictory — step 2 says 'Search the file using Grep or Read' while a later rule says 'Never Read the log file directly' — and the escalation path is stated twice rather than once, cleanly. | 4 / 5 |
Progressive Disclosure | The body is a genuine overview and both referenced files (references/source-maps.md, references/failure-scenarios.md) exist, are one level deep, and hold content appropriately split from SKILL.md (source-resolution internals and failure recovery). Navigation is easy; no nesting beyond one level. | 5 / 5 |
Total | 18 / 20 Passed |