Security audit
Security checks for vulnerabilities and agentic risk
This skill is a disclosed research briefing workflow for LLM inference acceleration, with no artifact-backed evidence of hidden execution, credential access, exfiltration, or destructive behavior.
Install this if you want a Chinese-first LLM inference acceleration research radar. Specify your preferred output language and confirm before allowing persistent knowledge-base updates if you do not want the skill to record durable research preferences or feedback.
The description says to use the skill for broad requests such as "daily or weekly briefs," "paper radar," "open-source watch," and "maintaining a research radar," which can overlap with many general research or summarization tasks. It does not provide exclusion conditions or negative examples to clarify when this skill should not activate, increasing the risk of unintended invocation.
The instruction "Default to Chinese output" imposes a language choice without indicating user consent or offering a language option. This is a natural-language locale policy concern because it may override the user's preferred language by default.
The workflow says to "generate a Chinese brief" by default unless the user specifies otherwise, which again enforces a language default without affirmative opt-in. This creates a policy issue because the skill should offer language choice rather than silently selecting a locale.
