memory-integration
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to retrieve and process data from previous conversation sessions and external knowledge graphs, which creates a surface for indirect prompt injection.
- Ingestion points: Data enters the context via
mcp__plugin_episodic-memory_episodic-memory__search,mcp__plugin_episodic-memory_episodic-memory__read, andmcp__memory__search_nodestools. - Boundary markers: The skill instructions do not specify the use of delimiters or clear separation between retrieved memory content and the agent's active instructions.
- Capability inventory: The skill uses tools capable of searching conversation history, reading JSONL files (e.g.,
/path/to/conversation.jsonl), and performing CRUD operations on a fact-based knowledge graph. - Sanitization: There are no documented mechanisms for sanitizing or escaping instructions that may be embedded within the retrieved episodic or fact-based memory.
Audit Metadata