Security audit
Security checks for vulnerabilities and agentic risk
This task-queue skill is mostly purpose-aligned, but it asks agents to run a required helper script that is not included and uses broad activation wording with persistent task data.
Review before installing. Use it only for explicit queue-management tasks, verify or supply the intended queue_task.py implementation yourself, and avoid placing secrets, tokens, regulated data, or unnecessary personal data in queued payloads unless you are comfortable with local persistence and any downstream API transfer.
The skill declares an overly broad activation scope using many generic workflow phrases, which increases the chance an agent will invoke it outside the user's intended context. Because the skill has exec capability and persists files, accidental triggering can cause unintended filesystem writes, queue creation, or task-state mutation without sufficiently explicit user intent.
The skill prominently advertises persistent local storage of task payloads and status but does not clearly warn that user-supplied data will be written to disk. In an agent context, users may provide sensitive content assuming ephemeral handling, leading to unintended local retention of confidential data in JSONL and state files.
L135 明确写出 queue.enqueue(..., priority=1) 并说明“支持优先级”，而 L465 又声明免费版“不支持基于优先级的动态调度”。这不是单纯信息缺失，而是同一技能文档对能力边界给出了互相冲突的描述，容易误导使用者对实际功能范围的判断。
The user-sync example encourages storing user profile data in queue files and transmitting it to another system without any privacy, minimization, or retention guidance. In practice this can expose personal data through local logs, archives, backups, or mishandled development environments, especially since the design uses append-only JSONL files.
The API batch-calling example sends task data to external services but omits any warning about outbound data transfer, third-party sharing, or service trust boundaries. In agent workflows, this can cause users to unknowingly export sensitive identifiers or business data to remote APIs, compounding the persistence risk already present in the local queue files.
该行明确将“完全中文化表达”作为改造内容之一，但文档未说明这只是默认语言，也未提供其他语言/locale 的选择。按照语言/locale 政策，强制单一语言且无用户 opt-in 可能构成自然语言策略违规。
