Security audit
Security checks for vulnerabilities and agentic risk
This appears to be a document-only AI governance skill, but it asks for broader write/network-capable tools than its stated purpose needs.
Install only if you are comfortable with a governance checklist skill whose declared tool access is broader than its stated text-only behavior. Prefer enforcing no network and no write/edit permissions, and treat the A³ review as advisory unless the trigger scope is tightened.
The file presents the skill documentation entirely in Chinese and directs human readers to another file, but it does not offer any language option or explain that the skill is intentionally region- or locale-specific. Under the policy, language constraints should be opt-in or clearly justified rather than implicitly imposed.
This markdown file defines invocation cues using a mixed list of trigger phrases, several of which are high-level concepts rather than narrow commands, such as '自我修改', '自主行动评估', and 'AI三定律'. Without explicit scope limits or negative examples, the activation boundary is unclear and may cause unintended invocation in ordinary policy or research discussions.
The manifest says the skill should be read when an AI is about to perform an irreversible or high-impact autonomous action, but it does not define what qualifies as "high-impact" in the trigger itself. This ambiguity could cause the skill to activate in many unrelated situations beyond its intended scope.
The trigger list mixes specific phrases with broader concepts such as "self-modification" and "trigger threshold," which may appear in general governance or technical discussions. Without explicit contextual limits or exclusion conditions, these phrases may cause unintended invocation collisions.
The usage text says to activate via keywords or auto-trigger the A³ review flow, but it does not define which keywords are authoritative or when auto-triggering should not occur. This makes activation behavior ambiguous and risks overly broad invocation.
The Chinese description states that users can activate the mechanism by keywords or the system can auto-trigger A³ review in the background, but it does not define boundaries for those behaviors. Missing exclusions or precise scope can lead to accidental activations during ordinary discussion.
The file content is entirely in Chinese and presents the skill index and trigger description only in that language. There is no indication that users or agents may choose another language, nor any justification that this skill is specifically limited to a Chinese-language or region-specific context.
The manifest uses Chinese-only natural-language values for the package and author fields (e.g. "未命名包" and "注册老炮") without any indication that the skill is region-specific or that users may choose another language/locale. Per the policy, fixed locale/language choices can be a natural-language policy issue when no opt-in or justification is provided.
