Security audit
Security checks for vulnerabilities and agentic risk
This is a narrow QA validation skill that reviews generated test cases for hallucinations and does not show hidden execution, persistence, exfiltration, or destructive behavior.
Install only if you want a Chinese-language QA workflow for validating generated test cases against requirements. Do not copy-paste the companion npx install command unless you trust the publisher and have checked the package or pinned source you intend to install.
The trigger phrases include broad requests like '验证一下输出', '检查有没有幻觉', and '确认一下质量', which can overlap with many ordinary review interactions. In an agent system, that ambiguity can cause unintended skill activation, pulling in the wrong workflow or context and potentially altering outputs, consuming extra tools/context, or interfering with safer/more appropriate review paths.
The manifest description in L004 is written as an instruction set entirely in Chinese and presents the skill behavior without indicating that language is configurable or optional. Under the policy, forcing a specific language/locale without user opt-in is a natural-language policy concern unless the locale restriction is explicitly justified.
The metadata says the skill auto-activates 'after AI generates test cases, before final output' and on vague user phrases, but it does not define firm boundaries for when auto-activation should be suppressed. That ambiguity increases the chance of accidental invocation, workflow hijacking, or recursive/duplicate validation in multi-skill environments, which can degrade reliability and create denial-of-service-like overhead.
The skill instructs users to run npx skills add Kokxi/qa-test-skills without pinning an exact package version or immutable source. That creates a supply-chain risk: a later compromised or maliciously updated package could be fetched and executed at install time, and the presence of installation guidance inside a skill increases the chance an operator will copy-paste it without additional verification.
SQP-3 applies to all file types and covers language or locale policy violations. This markdown file presents all operational guidance in Chinese only, and nowhere indicates that users may choose another language or that the skill is intentionally region-specific.
