Security audit
Security checks for vulnerabilities and agentic risk
This monitoring skill is broadly aligned with its purpose, but it asks for high-impact system, network, and credential access without enough scoping or user-control safeguards.
Review this skill carefully before installing. Use it only in a controlled environment, provide read-only or least-privilege monitoring credentials, avoid storing secrets in local config files unless protected, and require explicit confirmation before remote host access, log export, callbacks, webhooks, Docker socket access, or kubeconfig use.
The document gives conflicting secret-handling guidance: it says API keys are configured via environment variables, then states all credentials are stored in local configuration files. That ambiguity can lead operators or downstream agents to persist sensitive tokens in plaintext files, increasing the chance of credential disclosure through logs, backups, or repository leakage.
The description includes broad 'Use when' language around monitoring, log analysis, alerting, and deployment management, which creates very loose invocation boundaries. In an agent environment with exec capability, generic user requests may trigger this skill unexpectedly and lead to network access, system inspection, or command execution beyond the user's intended scope.
The README says natural-language instructions can directly trigger the tool without defining strict invocation boundaries or approval gates. Because the skill advertises read and exec tools plus multi-node monitoring features, this raises the risk of unintended execution, collection, or remote access from vague prompts.
The skill advertises export functionality but does not clearly warn that data may be written to disk. In a monitoring/log-analysis context, exported files can contain operationally sensitive information, causing unintended persistence of logs, host details, or alerts on the local filesystem.
The examples include callback URLs, Grafana URLs, webhooks, and remote monitoring actions but do not clearly disclose that data may leave the local environment or traverse the network. For infrastructure monitoring, this can expose topology, metrics, logs, and alert content to external systems without informed operator consent.
The documentation instructs use of a private SSH key path for adding monitored nodes, in a skill that exposes exec capability and broad natural-language triggering. This encourages handling raw private-key material during routine operations and increases the risk that an agent, script, or operator may expose, misuse, or improperly forward highly privileged credentials.
python3 （请参考skill目录中的脚本文件） add \
--name "web-01" --host 192.168.1.10 --ssh-key $HOME/.ssh/id_rsa
python3 （请参考skill目录中的脚本文件） add \
--name "db-01" --host 192.168.1.20 --ssh-key $HOME/.ssh/id_rsa
# ...
# 批量添加
python3 （请参考skill目录中的脚本文件） batch-add --file servers.csv
The explicit kubeconfig parameter reinforces that the skill may access Kubernetes authentication material as part of normal use. Because kubeconfig often grants broad cluster access, any accidental reading, logging, export, or reuse of that file can lead to unauthorized control over workloads and sensitive cluster data.
socket: "/var/run/docker.sock"
kubernetes:
enabled: true
kubeconfig: "$HOME/.kube/config"
# ...
logs:
aggregation: true
The explicit kubeconfig parameter reinforces that the skill may access Kubernetes authentication material as part of normal use. Because kubeconfig often grants broad cluster access, any accidental reading, logging, export, or reuse of that file can lead to unauthorized control over workloads and sensitive cluster data.
socket: "/var/run/docker.sock"
kubernetes:
enabled: true
kubeconfig: "$HOME/.kube/config"
# ...
logs:
aggregation: true
No suspicious patterns detected.