Security audit
Security checks for vulnerabilities and agentic risk
This is a disclosed local caching and request-learning skill, with privacy considerations around stored query logs but no hidden or unsafe behavior in the inspected artifact.
Before installing, understand that this skill is designed to cache results and learn from repeated requests, so query text and usage patterns may be stored locally in cache and log files. Avoid using it for secrets, keys, passwords, personal data, or highly sensitive business prompts unless you have reviewed the implementation and retention behavior.
The skill advertises automatic learning and recommendation features based on repeated requests, which implies retention and processing of user request content over time. In the context of an agent skill, users may submit secrets, financial data, or personal information, so storing raw requests creates a meaningful confidentiality risk even if the stated purpose is optimization.
The examples show learner.record_request(query=...) capturing raw query text and later generating recommendations from those records. This is dangerous because user-provided content may contain sensitive business data or personal information that can later be exposed through logs, summaries, recommendations, or exported reports.
The documentation explicitly names persistent files like patterns.json and request_log.jsonl, but it does not clearly warn that user queries and behavioral data may be stored on disk. In a skill that learns from request content, this creates a real privacy and data-retention risk because sensitive prompts may be written to local storage without informed user consent or minimization.
The documented file structure includes persistent request-pattern and request-log files, confirming that learned data is retained beyond runtime. This makes the risk more concrete because any local compromise, shared workspace access, backups, or accidental disclosure can expose historical user inputs and inferred behavior.
The natural-language content of the skill is presented entirely in Chinese, and there is no indication that users may choose another language or that the skill is intentionally limited to a Chinese-speaking or region-specific audience. This can violate a language/locale policy when a skill implicitly forces a specific language without opt-in.
