SkillPulse

Profile offline OpenClaw skill run samples to detect latency, CPU, and memory bottlenecks and compare sessions.

Security Scan

This is an offline Node.js CLI that reads JSON sample files, computes simple profiling metrics, and writes reports — its code, declared requirements, and runtime instructions are consistent with that purpose.

Purpose & Capability

Name/description (profile offline samples for latency/CPU/memory) matches the code and SKILL.md. The skill only requires node and works on local JSON samples. No unrelated credentials, binaries, or system config are requested.

Instruction Scope

SKILL.md instructs running the bundled CLI against local JSON sample files and producing JSON/HTML reports. The runtime instructions and source only read the provided input files and write report files; they do not access other system paths, network endpoints, or undeclared environment variables.

Install Mechanism

No install spec is provided (instruction-only with code files). Files are included in the package; there is no external download, no third‑party install script, and package.json has no runtime dependencies, so nothing is pulled from remote at install time.

Credentials

The skill declares no required env vars or credentials and the source code does not read environment secrets. It only relies on the 'node' binary and user-supplied file paths, which is proportionate to the stated functionality.

Persistence & Privilege

always is false and the skill does not attempt to modify other skills or system-wide agent settings. It writes report files to paths specified by the user (expected behavior for a CLI).

Assessment

This skill appears to be a simple, local Node.js CLI for analyzing JSON sample logs. Before installing or using it: (1) ensure you have Node.js >=18; (2) only run it on sample files you trust or have sanitized — the tool will read any JSON file path you provide and could include sensitive data in generated reports; (3) review or run the included tests (test.js) in an isolated environment if you plan to run it in CI; and (4) because it writes files to user-specified paths, avoid supplying output locations that would overwrite important files. There are no network calls or secret accesses in the code, so the risk of remote exfiltration appears low based on the provided source.

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.