Security audit
Security checks for vulnerabilities and agentic risk
This skill is a disclosed research-paper triage rubric with web-search guidance, not a skill that installs code, persists, or accesses private credentials.
Install this if you want a very strict, Chinese-language research-paper reviewer. Avoid using it on confidential manuscripts unless you are comfortable with the agent performing external searches based on the paper’s title, methods, or baselines.
The entire operative instruction set is written in Chinese and directs the agent's role, workflow, and output format exclusively in that language. There is no indication that users may choose another language or that the Chinese-only constraint is required for a region-specific or compliance-specific purpose, which makes this a natural-language locale policy issue.
The manifest describes a paper-triage skill focused on methodological novelty and explicitly says to ignore author, institution, venue, and citation prestige. However, the documented behavior expands into mandatory external related-work and ecosystem investigation, including follow-up papers, implementation searches, version tracking, and industrial/open-source adoption checks, which goes beyond strict paper-intrinsic triage. This is a semantic expansion of scope rather than a mere implementation detail.
The skill enables implicit invocation for a broad paper-triage capability without any narrowly scoped activation condition. That creates a prompt-surface where the agent may auto-select this skill in unrelated contexts, causing unintended evaluation behavior, leakage of user content into the skill workflow, or policy-bypassing routing decisions based on an overbroad trigger.
No suspicious patterns detected.
