System Load Monitor
PassAudited by ClawScan on May 1, 2026.
Overview
This looks like a simple local system-load checker; users should know it may delay agent work, expose process command details, and mentions optional Feishu alerts.
This skill appears safe to use for local load checks. Before installing, be comfortable with the agent delaying work under high load, avoid sharing full process-list output if command lines may contain secrets, and approve any Feishu notification behavior explicitly.
Findings (3)
Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.
The agent may pause or defer resource-intensive work when the system is busy.
The skill explicitly changes task continuation behavior based on load status. This is central to the stated purpose, but it can delay user-requested tasks.
- **critical**: Pause the task and retry after cooling down
Use this skill when task pacing is desired, and adjust thresholds so it does not pause work too aggressively.
Process names and command-line arguments may be shown to the agent or included in copied diagnostic output.
The script returns top process command lines as part of its diagnostic output. That is useful for monitoring, but command lines can contain sensitive arguments or untrusted text.
"command": ' '.join(parts[10:])
Treat process-list output as local diagnostic data, avoid sharing full output externally, and redact any command arguments that contain secrets.
If the agent has a Feishu integration, system-load alert details could be sent outside the local environment.
The instructions mention sending alerts through an external messaging provider, but the artifacts do not define recipient, credentials, or notification contents. This appears to be a purpose-aligned alert suggestion rather than implemented exfiltration.
Notify the user (via Feishu message)
Require explicit user approval before sending Feishu messages and keep alert content limited to what the user expects.