中文 集群监控

AdvisoryAudited by Static analysis on May 8, 2026.

Overview

No suspicious patterns detected.

Findings (0)

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.

What this means

The agent could disrupt services, kill the wrong process, or remove data while trying to fix a cluster issue.

Why it was flagged

The instructions move beyond passive monitoring and tell the agent to attempt repairs, terminate processes, and delete old data when conditions are detected, without requiring explicit user approval or defining safe limits.

Skill content
如果检测到服务关闭:
1. 记录问题到 memory.md
2. 尝试重启服务
...
| 端口被占用 | 使用 `lsof` 查找并终止进程 |
| 磁盘已满 | 清理日志,删除旧数据 |
Recommendation

Require explicit user confirmation before any restart, process termination, SSH tunnel change, log cleanup, or data deletion, and limit actions to named services and paths.

What this means

Installing this skill could lead the agent to request or use elevated privileges to change system service state.

Why it was flagged

These commands require administrator-level host privileges, but the skill does not specify when to ask the user, how to limit privilege use, or how to verify the target services safely.

Skill content
# 重启 PKM API
sudo systemctl restart pkm-api

# 重启 Qdrant
sudo systemctl restart qdrant
Recommendation

Treat sudo/systemctl actions as manual, user-approved operations only, and document the exact services, hosts, and rollback steps.

What this means

Operational details about the user's local cluster could be shared outside the conversation unexpectedly.

Why it was flagged

The skill directs notification to a named recipient without defining who that person is, what channel to use, what information is shared, or whether the current user must approve it.

Skill content
3. 如果重启失败,通知 Alexandre
Recommendation

Remove the automatic notification step or require explicit user approval with a disclosed recipient, channel, and message content.

What this means

Cluster status, service names, and troubleshooting history may persist across future tasks.

Why it was flagged

The skill persists incident and repair information to memory.md, which is purpose-aligned for operations logging but lacks a defined path, retention policy, and reuse boundary.

Skill content
1. 记录问题到 memory.md
...
4. 记录修复尝试
Recommendation

Keep any memory.md log scoped to this cluster, avoid secrets or sensitive incident details, and give users a way to review and delete it.

What this means

Users may not realize the skill expects local command execution tools and administrative service controls.

Why it was flagged

The instruction-only skill depends on several local binaries and privileged utilities, while the registry metadata declares no required binaries; this is an under-declared operational surface rather than hidden code.

Skill content
curl -s http://localhost:8001/health ... python3 -c ... sudo systemctl restart pkm-api ... openclaw gateway restart
Recommendation

Declare the expected binaries and clarify which commands are read-only checks versus privileged maintenance actions.