ac-rules-expert

Warn

Audited by Gen Agent Trust Hub on Feb 18, 2026

Risk Level: MEDIUMREMOTE_CODE_EXECUTIONDATA_EXFILTRATION
Full Analysis
  • Dynamic Execution (MEDIUM): The skill documents the use of safe_exec() for evaluating Python-based filters in references/core-components.md. This creates a runtime environment for executing user-defined scripts, which is a high-risk surface for potential sandbox escapes or resource exhaustion.
  • Data Exposure & Exfiltration (MEDIUM): In references/core-components.md and references/rule-evaluation.md, multiple code snippets demonstrate building SQL queries via f-strings with direct variable interpolation (e.g., WHERE parent = '{user}'). These patterns are susceptible to SQL injection if the variables are not strictly sanitized, potentially allowing attackers to bypass access controls or exfiltrate database content.
  • Privilege Escalation (MEDIUM): The 'Query Filters Report' described in references/debugging-reports.md includes an impersonate_user capability. This feature allows a user to trigger backend logic in the context of another user, which could be abused to perform reconnaissance on administrative accounts or bypass record-level permissions.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Feb 18, 2026, 12:06 AM