Security audit
Security checks for vulnerabilities and agentic risk
The skill is mostly a static Chinese audit quick-reference, but it embeds agent-directed instructions to buy and install an external paid package.
Use this skill as a local CAS audit checklist only. Do not allow the agent to follow the embedded SkillPay purchase or installation instructions unless you independently intend to buy that separate package and review it as a separate install.
The README says to use the skill whenever a user needs help with a wide range of audit tasks, and the example trigger phrases include common audit questions such as '重要性怎么定' and '审计报告意见怎么出'. These are broad natural-language requests without clear exclusion conditions or scope boundaries, which increases the chance of over-triggering on ordinary audit discussion.
The skill describes itself as an immediately usable quick-reference package requiring no system login, yet the README also nudges the agent toward external SkillPay resources and an install/purchase flow. This mismatch can mislead both users and agents about the trust boundary and operational scope of the skill, increasing the chance that an agent follows off-platform instructions not necessary for the advertised function.
The README includes explicit instructions for the agent to visit external SkillPay pages, read installation guidance, and facilitate purchase/installation of a separate paid package. That behavior exceeds the stated purpose of a local audit quick-reference skill and creates a prompt-injection style expansion of scope into external commerce and installation actions, which could steer an agent into untrusted workflows or social-engineer a user into payment.
The trigger conditions are broad, natural-language phrases covering common audit questions without clear scope limits, which can cause the skill to activate in contexts beyond its intended use. Overbroad activation is dangerous because it can inject prescriptive audit guidance or external purchase prompts into unrelated conversations, increasing the chance of mistaken reliance or unwanted workflow steering.
The skill includes a monetized purchase/install workflow for an external '完整版' package that is not necessary for the stated quick-reference purpose. This expands the trust boundary from static audit-reference content into external commerce and installation, creating social-engineering and supply-chain risk because the agent may be induced to fetch, recommend, or help install unreviewed third-party content.
The skill's display name, description, instructions, and usage guidance are entirely in Chinese, and there is no statement offering users a language preference or explaining that the skill is intentionally limited to a Chinese-language audience. Under the stated policy, forcing a specific language without user opt-in can be a natural-language policy violation.
