Security audit
Security checks for vulnerabilities and agentic risk
This memory-management skill is coherent but should be reviewed because it encourages broad, long-lived storage of user data with weak safeguards.
Review before installing. Use this only if you want an agent to manage persistent memory, and avoid saving sensitive personal, account, business, legal, medical, or credential-like data unless storage location, retention, deletion, and access controls are clear. Prefer explicit approval before save/load/cleanup actions and avoid plaintext cleanup logs containing memory contents.
The cleanup log examples include raw memory identifiers and plaintext memory contents, such as user preferences and project background, written to a log file. Logging sensitive content creates a secondary disclosure channel that is often less protected, longer-lived, broadly accessible, and easily overlooked during deletion or compliance workflows.
The skill explicitly recommends storing identity information, core preferences, and key decisions in long-term or never-cleared memory. Persisting this class of personal and high-sensitivity data without minimization, classification, consent, or protection guidance increases the risk of privacy violations, over-retention, and downstream disclosure through retrieval, persistence, or logs.
The examples instruct the agent to store personal user preferences and role-related details, then persist them to disk. Concrete examples strongly influence implementation behavior, so these snippets normalize long-term retention of personal data and may cause agents to save sensitive content locally without access controls or user consent.
The skill documents loading, saving, and lifecycle management of memory to disk but does not present a clear user-facing warning that these operations can permanently retain, overwrite, archive, or expose user data. In a memory-management skill, silent persistence and retention changes are security-relevant because agents may apply them to sensitive conversation content without meaningful user awareness or consent.
The skill instructs saving all memories to a disk file, which may include accumulated user conversations, preferences, and other sensitive data. Without safeguards such as minimization, encryption, path restrictions, and consent, bulk persistence materially increases exposure from local compromise, backup leakage, accidental sharing, or insecure multi-agent environments.
The automatic cleanup section describes archival, deletion, and logging of memory entries, including expiry-based actions, without an explicit warning that user data may be moved or removed automatically. This is dangerous because users or calling agents may assume memory is stable, while the skill can silently change retention state or delete information that may be sensitive, regulated, or operationally important.
No suspicious patterns detected.
