ai-brain
Pass
Audited by Gen Agent Trust Hub on Jun 24, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill's primary function is to manage a local memory store using standard filesystem tools. The behavior aligns with the stated purpose of creating a persistent knowledge base in an Obsidian vault.
- [DATA_EXPOSURE]: The skill instructions specify that the agent should read and write data to a local directory ('AI Brain/'). It includes explicit safety guidelines for the agent to respect the privacy of the user's vault and to restrict its operations to the specific memory folder.
- [COMMAND_EXECUTION]: The skill utilizes the Filesystem MCP to perform file and directory operations such as creating the memory folder, writing session summaries, and searching through existing notes. These operations are within the scope of the intended functionality.
- [PROMPT_INJECTION]: As the skill retrieves context from previously saved conversations, it possesses an indirect prompt injection surface. Malicious instructions provided by a user (or contained in content the user asks the agent to remember) could be stored and later interpreted as instructions during the 'Recall' phase. This is an inherent risk of long-term memory systems but is not a direct flaw in the skill's code.
Audit Metadata