Security audit
Security checks for vulnerabilities and agentic risk
This memory skill is not malicious, but it asks the agent to persist and update user-related memory automatically without clear consent or retention controls.
Install only if you want an agent to keep long-term local memories and update them as part of normal use. Before using it with sensitive work, set explicit rules for when memory may be read or written, require review before saving session summaries, and define how stored memories can be inspected, edited, or deleted.
The trigger list contains broad phrases like '长期记忆', '记忆管理', and '总结' that are likely to appear in ordinary conversation, so the skill could activate unintentionally. Because the skill performs persistent memory and automated workflows, accidental invocation can cause data retention or other side effects without explicit user intent.
The skill explicitly advertises cross-session persistent memory, structured storage of user-related facts, and automatic memory management. Without clear data minimization, consent, retention limits, and review controls, this creates privacy and profiling risks by retaining user information beyond the current session.
The skill states that it will automatically execute boot sync, read structured data, and scan transaction-day fund flows, which are proactive actions with no clear user warning or approval boundary. Automatic scanning and data access increase the chance of over-collection, privacy violations, or unintended actions being taken on session start.
The wrap-up workflow instructs the system to update changelogs, active documents, and memory at the end of each session, but it does not warn the user that persistent data will be modified. This creates a risk of silent retention and mutation of stored information, including potentially sensitive conversation content.
The session closeout flow mandates updating memory after each conversation, which means session content may be retained automatically even when the user has not asked for persistence. In the context of a memory skill, this makes the behavior more dangerous because retention is a core feature and is coupled to routine session termination, increasing the likelihood of silent collection.
The skill description is written entirely in Chinese, which implies a fixed language presentation without offering any user language or locale choice. This can violate language/locale policy when users are not given an explicit opt-in or alternative.
