Content
75%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 highly actionable, well-sequenced diagnostic skill with concrete commands across all environment/symptom combinations and real reference-file separation. Its main weaknesses are repetition of the trigger and dump-deferral guidance across sections, and inlining of detailed command content that the reference files are meant to hold.
Suggestions
Consolidate the PerfView /StopOn trigger guidance and the 'invoke dump-collect, do not provide dump commands directly' deferral into single canonical sections; reference them from Step 2 and Common Pitfalls instead of restating.
Move the per-symptom detailed command blocks (Slow Requests, Networking, GC, Assembly Loading) into the corresponding reference files, keeping the SKILL.md body to triage rules and a pointer per environment.
Tighten the Validation section into an explicit feedback loop (e.g., 'if the collection command errors, verify PID and privileges, then re-run') so error recovery is part of the workflow rather than a separate checklist.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Mostly efficient and full of genuinely specialized domain knowledge (provider keywords, version-specific flags, kernel prerequisites), but the same guidance is repeated in multiple places — the PerfView /StopOn trigger warning appears in Step 2, Common Pitfalls, and the long-repro section, and the 'do not provide dump commands directly, invoke dump-collect' deferral is restated several times. | 3 / 5 |
Actionability | Provides copy-paste-ready, executable commands for every environment and symptom combination — e.g. 'PerfView /ThreadTime collect /BufferSizeMB:1024 /CircularMB:2048', 'dotnet-trace collect -p <PID> --profile gc-verbose', 'dotnet-trace collect-linux --profile thread-time', and concrete provider specs like 'Microsoft-Windows-DotNETRuntime:0x4:4'. | 5 / 5 |
Workflow Clarity | A clear four-step sequence (understand environment → recommend tools → guide collection → recommend analysis) with a PID-discovery verification checkpoint, a Validation checklist, an artifact-handoff checklist, and a Common Pitfalls troubleshooting table, but the validation items are post-hoc checks rather than an integrated 'if fail then fix and retry' feedback loop. | 4 / 5 |
Progressive Disclosure | Good structure with a tool-reference-lookup table mapping each environment to one-level-deep reference files (all five referenced files exist) and clear inline links, but the body also inlines substantial detailed per-symptom command content that overlaps with the 'detailed command-line usage' promised to live in the reference files. | 4 / 5 |
Total | 16 / 20 Passed |