Security audit
Security checks for vulnerabilities and agentic risk
This skill is a local design-preference helper with disclosed preference-file persistence, but users should understand it may automatically record and reuse design choices.
Install this only if you want an agent to maintain a reusable local profile of your visual preferences. Review ~/.design-preferences/profile.md periodically, remove entries you do not want reused, and avoid storing sensitive personal or client information in that profile.
The skill states that it will automatically learn preferences and store them in local files, but it does not provide a clear consent boundary or warning before creating or modifying user data. In an agent context with exec access, silent writes to persistent files can surprise users, overwrite existing content, or accumulate sensitive preference/profile data without explicit approval.
The capability keywords are excessively broad and ambiguous, effectively positioning the skill to match many generic design-related prompts. That broad scope is risky because it can route unrelated tasks into a skill that performs persistent state changes, increasing unintended data collection and modification risk.
This duplicated finding points to the same persistent-write behavior: initialization commands write user preference data to ~/.design-preferences/profile.md. In context, the danger is not remote code execution but undisclosed long-term local state that may collect sensitive preference information and affect later sessions without clear consent.
# 初始化偏好文件
mkdir -p ~/.design-preferences
cat > ~/.design-preferences/profile.md << 'EOF'
# 我的视觉偏好
This duplicated finding points to the same persistent-write behavior: initialization commands write user preference data to ~/.design-preferences/profile.md. In context, the danger is not remote code execution but undisclosed long-term local state that may collect sensitive preference information and affect later sessions without clear consent.
# 初始化偏好文件
mkdir -p ~/.design-preferences
cat > ~/.design-preferences/profile.md << 'EOF'
# 我的视觉偏好
The trigger condition covers broad everyday design activities such as design creation, UI design, poster work, and brand visuals, which can cause the skill to activate in ordinary requests where the user did not intend persistent preference tracking. In a tool-enabled agent, overbroad activation increases the chance of unsolicited file reads/writes and profile accumulation across unrelated tasks.
The description and main documentation emphasize a lightweight local tool where 'all preference data is stored locally' and 'no extra API key' is needed, implying a local preference-tracking utility. However, the dependency and limitation sections state that an LLM API is required and that operation depends on network conditions, which expands the behavior beyond a purely local Markdown preference store.
The manifest description and the rest of the skill are presented entirely in Chinese, with no indication that users may choose another language or that the skill is intended only for a Chinese-language context. This can violate language or locale choice expectations under the policy.
The skill is presented as a personal visual preference learning tool for UI and graphic design. Documenting 'ping' based network testing introduces a troubleshooting capability unrelated to design preference management and not justified by the stated purpose.