Security audit
Security checks for vulnerabilities and agentic risk
The skill is a database scheduler helper, but it can create or remove scheduled jobs, run jobs immediately, and start or stop a background scheduler without clear confirmation or safety boundaries.
Review this skill before installing if you use production or shared databases. It should be used only when you want the agent to manage database backups, scheduled jobs, workflows, and the scheduler daemon; require explicit confirmation for task deletion, immediate runs, daemon start/stop, and large backups.
The manifest front matter advertises only backup, task, and log capabilities, but the body also instructs daemon lifecycle control and workflow orchestration. This mismatch can cause downstream policy, approval, or routing systems to under-classify the skill's privileges, leading an agent to expose or invoke more powerful operations than expected.
The trigger phrase '定时任务' is broad and can match many benign scheduling discussions, causing the skill to activate when the user did not intend concrete task management. In a skill that can add, remove, run tasks, or affect a scheduler daemon, over-broad activation raises the risk of unintended operational actions.
The skill activates on listed keywords without clear scope or confirmation boundaries, despite including operational commands with side effects. Broad keyword routing can cause accidental invocation in ambiguous conversations, which is more dangerous here because the skill can change scheduler state and execute jobs.
The skill documents destructive or system-affecting operations such as task removal and daemon start/stop without warning or confirmation guidance. In practice, this can lead an agent to disable automation, interrupt scheduled jobs, or remove tasks without the user appreciating the operational consequences.
The documented use scenarios imply the skill is limited to backup/task/log routing, but later sections instruct handling daemon and workflow commands too. This inconsistency increases the chance of unintended agent behavior, especially if operators rely on the top summary to understand what the skill is allowed to do.
Backup operations can write substantial data, consume disk space, and potentially affect system performance, but the skill gives no warning about those effects. This omission is less severe than direct destructive actions, yet it can still cause avoidable outages or storage exhaustion in constrained environments.
No suspicious patterns detected.
