Skill Profiler

Analyze offline sample sets and surface the slowest or heaviest skills before they become production regressions.

When to use

• You have JSON sample logs with latencyMs, cpuMs, and memoryMb.
• You want threshold-based hotspot detection for local review or CI.
• You need a shareable JSON or HTML report, or a before/after comparison between two runs.

Commands

node {baseDir}/bin/skill-profiler.js run --input ./samples.json
node {baseDir}/bin/skill-profiler.js report --input ./samples.json --out ./report.html
node {baseDir}/bin/skill-profiler.js compare --left ./v1.json --right ./v2.json

Sample input

[
  {
    "sessionId": "s1",
    "skill": "clawshield",
    "latencyMs": 1320,
    "cpuMs": 910,
    "memoryMb": 240
  }
]

What it reports

• Average latency and p95 latency per skill
• Average CPU and peak memory per skill
• Bottlenecks using configurable latency, CPU, and memory thresholds
• Session diffs showing added, removed, and changed skills

Boundaries

• Skill Profiler is built for offline sample analysis, not live tracing.
• It depends on the quality of the input samples and does not capture traces by itself.