Security audit
Security checks for vulnerabilities and agentic risk
This study assistant is purpose-aligned and transparent about storing notes, using web search, and creating review reminders.
Install only if you want a study assistant that writes structured files into an Obsidian vault, updates indexes, may send search terms to web search, and may create review reminders. Use explicit prompts for file-changing actions and avoid putting sensitive drafts or private research topics into web-search requests unless you are comfortable with that data flow.
The activation rule says the skill triggers whenever the user says any listed phrase, without contextual constraints or confirmation. In a skill that can write files, update indexes, search externally, and create reminders, ambiguous triggering increases the chance of unintended actions and data handling.
The activation rule says the skill triggers whenever the user says any listed phrase, without contextual constraints or confirmation. In a skill that can write files, update indexes, search externally, and create reminders, ambiguous triggering increases the chance of unintended actions and data handling.
The skill persists books, notes, exam-prep material, and drafts in a local Obsidian vault without warning that potentially sensitive study materials, personal notes, or drafted documents will be stored on disk. This is dangerous because users may assume the interaction is ephemeral while the skill retains structured content in predictable locations.
The skill describes automatic updates to MOC.md whenever notes are created or deleted, but it does not clearly warn the user that local files will be modified as a side effect. Hidden or implicit file modification is risky because users may not realize the skill is changing persistent project state beyond the requested note action.
The WebSearch-based features and academic tracking imply that user queries may be sent to external services, but the skill does not warn the user about this data flow. Search terms can contain sensitive research interests, private project names, or draft content, creating an avoidable privacy risk.
The skill describes creating scheduled reminders via CronCreate without warning the user that a persistent background schedule will be established. Even if low severity, silent scheduling can surprise users, create notification spam, or leave behind ongoing automation they did not intend.
The instruction to address the user as “主人” imposes a specific form of address and tone without opt-in. This can conflict with organizational language or locale expectations because the user is not given a choice of preferred address style.
No suspicious patterns detected.
