Security audit
Security checks for vulnerabilities and agentic risk
This is a planning and coding-handoff skill with no executable payload; the main caveats are broad activation and fixed Chinese/English output defaults.
Install this if you want a PM-style planning layer that may read relevant repo files before producing specs and coding-agent handoffs. Be aware that it is tailored to an operator named Alan and defaults to Chinese user-facing plans plus English agent handoffs; users who need another language should override that explicitly.
The skill description is broad enough to capture many ordinary planning, brainstorming, and coding-prep requests, which can cause inappropriate auto-selection over narrower, safer, or more suitable skills. In an agent-routing system, this can degrade least-privilege behavior and increase the chance that a repo-reading, high-authority planning skill is invoked unnecessarily.
Mandating Chinese responses without user choice can cause instruction mismatch, misunderstandings, or reduced usability when the user expects another language. In security-sensitive workflows, forced locale changes can also obscure review and validation by downstream users or operators who cannot easily verify the output.
The file includes multiple default templates labeled and structured in Chinese, while the surrounding instruction says to use these templates as defaults. This can force a specific language/locale on users without explicit opt-in, which is a natural-language policy concern under the language/locale rule.
This JSON eval file contains natural-language expectations that require the skill to produce output in Chinese and English. Because the policy applies to all file types, embedding a fixed bilingual output requirement without explicit user choice or a documented justified locale constraint is a language-policy issue.
The prompt and expected output prescribe Chinese for the reviewable plan and English for downstream agent handoffs. This forces specific languages as part of the skill behavior and does not show user opt-in alternatives or a documented region-specific justification.
No suspicious patterns detected.
