ReasoningBank Intelligence

Pass

Audited by Gen Agent Trust Hub on Mar 4, 2026

Risk Level: SAFE
Full Analysis
  • [PROMPT_INJECTION]: No evidence of prompt injection attempts, safety filter bypasses, or instruction override patterns were found in the skill metadata or implementation snippets.
  • [DATA_EXFILTRATION]: The skill uses local persistence for its learning database (e.g., './reasoning-bank.db'). No network operations or attempts to access sensitive system files were detected.
  • [REMOTE_CODE_EXECUTION]: There are no remote code execution patterns, such as fetching scripts from external URLs or executing shell commands via subprocesses.
  • [COMMAND_EXECUTION]: The provided TypeScript examples demonstrate standard API interactions for recording and retrieving data. No arbitrary command execution or shell injection vectors are present.
  • [SAFE]: The skill relies on recognized dependencies within the agent's ecosystem (agentic-flow and AgentDB) for its primary functionality.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 4, 2026, 07:38 PM