memory
Pass
Audited by Gen Agent Trust Hub on Sep 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill establishes a workflow for reading and writing to a persistent knowledge store ($MEMORY_DIR) which may be updated by external synchronization mechanisms (referred to as "radio"). This creates a potential surface for indirect prompt injection if the synchronized data contains malicious instructions meant to influence the agent's behavior during its orientation phase.
- Ingestion points: The $MEMORY_DIR directory and data arriving via external "radio" synchronization.
- Boundary markers: The instructions do not define specific delimiters, XML tags, or "ignore embedded instructions" warnings for the agent when processing this memory.
- Capability inventory: The agent uses the
wikitool to read, write, update, and map files within the knowledge store. - Sanitization: No explicit validation or filtering logic is described for data arriving from external sync sources.
Audit Metadata