Robot
Pass
Audited by Gen Agent Trust Hub on Sep 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill defines a structured local memory system in
~/Clawic/data/robot/and requires the agent to read these files (such asmemory.mdandinventory.md) before making recommendations. This creates a surface where information provided by the user and stored in these files could contain instructions that influence the agent's behavior during future interactions.\n - Ingestion points: The agent is instructed to read
memory.md,inventory.md,projects/*.md, andcorrections.mdfrom the local data directory.\n - Boundary markers: There are no explicit delimiters or system instructions defined to differentiate between data and commands within the memory files.\n
- Capability inventory: The skill instructions suggest the use of file system operations (create directories, write files) and shell commands for robot package management and build processes (e.g.,
ros2,colcon).\n - Sanitization: The skill does not implement validation or sanitization for the data it retrieves from its persistent storage.
Audit Metadata