memory-bank

Pass

Audited by Gen Agent Trust Hub on Mar 11, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: Indirect prompt injection surface via persistent storage. \n
  • Ingestion points: The agent reads data from files located in the ./memory/ directory (SKILL.md). \n
  • Boundary markers: While the skill uses YAML frontmatter, it lacks explicit boundary markers or system instructions to disregard commands found within the body of memory files. \n
  • Capability inventory: The agent can execute shell commands (ls, head, rg) and use native file tools to read and write content. \n
  • Sanitization: There is no evidence of content sanitization or filtering before storage or after retrieval.\n- [COMMAND_EXECUTION]: Usage of shell commands for file management and searching. \n
  • The skill explicitly directs the agent to use ls, head, and rg (ripgrep) to interact with memory files. \n
  • These commands are restricted to a local directory, but they represent an active command execution capability.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 11, 2026, 09:10 PM