Security audit
Security checks for vulnerabilities and agentic risk
This is a broad Chinese-language task-planning skill that is mostly a disclosed methodology, with no executable payload or hidden persistence found.
Install this if you want a Chinese-language framework for decomposing and orchestrating tasks. Review the broad trigger terms first, and keep normal approval prompts enabled for file changes, code execution, browser automation, purchases, bookings, or other real-world side effects.
The trigger list is extremely broad and overlaps with common user vocabulary such as 'workflow', '任务分解', '执行框架', and 'Pipeline'. This creates a prompt-collision risk where the skill may activate unintentionally in unrelated conversations, causing hidden behavioral overrides and making downstream tool-using behavior more likely without clear user intent.
The trigger phrases are extremely broad and generic, including common terms like '工作流', 'Pipeline', '任务分解', and '执行框架', which are likely to appear in many ordinary user requests. This can cause the skill to activate unintentionally, hijack unrelated conversations, and insert its instructions into contexts where they were not requested.
The skill metadata and content are written entirely in Chinese and do not indicate any language negotiation or locale limitation. In a multi-language environment, this can lead to confusing or inaccessible behavior if the skill activates for users who did not request Chinese output, especially when paired with the broad trigger set.
The skill marks personal-life tasks as preferring full automation, and the accompanying examples include booking-related execution steps. In context, this encourages autonomous real-world actions on behalf of a user without an explicit requirement for confirmation, which can lead to unintended purchases, reservations, privacy exposure, or other external commitments.
The tool-calling section explicitly authorizes file writes, code execution, browser automation, and other external operations as interchangeable implementation details, but it does not require an explicit safety gate before system-impacting actions. Treating such actions as ordinary internal 'P' steps can hide risky side effects from users and increase the chance of unauthorized or insufficiently reviewed execution.
No suspicious patterns detected.
