Content
73%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.
The body is a well-written orchestrator with an exemplary gated workflow, concrete thresholds, and lean tables. Its fatal flaw is that the bundle does not back it: all rule files and the report template referenced by every phase are missing, so the skill's detailed methodology is unreachable as shipped.
Suggestions
Add the eight missing bundle files (rules/input-detection.md, rules/measurement-methodology.md, rules/react-profile-analysis.md, rules/chrome-trace-analysis.md, rules/heap-snapshot-analysis.md, rules/optimization-playbook.md, rules/confidence-loop.md, templates/analysis-report.md) or inline their essential content — the Workflow, Required-reading, and Memory-quickstart sections currently point at files that don't exist.
Merge the Workflow table and the "Required reading by phase" table into one table with columns for rule file and gate; the same phase-to-file mapping is currently listed three times (also in the confidence section).
For the CPU path, add one inline fallback command (like the heap-summary/heap-diff commands) so profile analysis is not entirely blocked by the missing rule files — e.g., a jq or node one-liner for top self-time functions from a trace.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense and assumes Claude's competence — no basic-concept explanations, tables for input formats and phases, concrete anti-pattern one-liners. Not a 5 because the phase-to-rule-file mapping is repeated across three sections (Workflow table, Required-reading table, Confidence-gated iteration), which could be consolidated. | 4 / 5 |
Actionability | Gives concrete executable commands for the memory path ("node --max-old-space-size=4096 <skill_dir>/scripts/heap-summary.mjs <snapshot.heapsnapshot>"), a concrete skill invocation (Skill(skill="confidence", args="analysis")), and actionable detection signals ("Top-level traceEvents array"). Not a 5 because the CPU-path analysis steps are delegated to rule files that don't exist in the bundle, so the guidance stops at pointers there. | 4 / 5 |
Workflow Clarity | Six clearly sequenced phases each with an explicit gate ("Do not skip a gate"), a validation feedback loop with concrete thresholds ("≥ 90%" proceed, "70–89%" one deeper pass, "< 70%" surface to user, max two deep-dives), and a Definition of Done checklist. This matches the anchor-5 example of sequence + explicit validation + feedback loops + checklist. | 5 / 5 |
Progressive Disclosure | Signaling is excellent — every phase maps to named rule files with "Load on demand — do not preload" — but scoring against the actual bundle, 8 of the 10 referenced files (all of rules/*.md and templates/analysis-report.md, which carry phases 0–5) do not exist. The thin-orchestrator structure is not actually realized: navigation leads to dead ends for the core content, which is neither inline nor in the bundle. | 2 / 5 |
Total | 15 / 20 Passed |