detecting-entra-offensive-tools-in-graph-logs

Pass

Audited by Gen Agent Trust Hub on Aug 7, 2026

Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The script scripts/agent.py performs network requests using the standard Python urllib library to interact with the Microsoft Azure Monitor Query API at api.loganalytics.io. This is a standard and expected operation for the skill's purpose of analyzing cloud logs.
  • [DATA_EXFILTRATION]: The authentication process involves transmitting a user-provided Azure Active Directory bearer token to the well-known Microsoft service domain api.loganalytics.io. This behavior is consistent with legitimate administrative workflows for querying Log Analytics workspaces and does not involve exfiltration to unknown or untrusted destinations.
  • [PROMPT_INJECTION]: The skill identifies a surface for indirect prompt injection as it retrieves and processes log entries that may contain attacker-influenced strings. Evidence chain: (1) Ingestion points: Workspace query results from api.loganalytics.io in scripts/agent.py; (2) Boundary markers: Absent as the data is used for display purposes; (3) Capability inventory: Network POST requests to a known API and console output; (4) Sanitization: Basic input validation using argparse type checking for numerical parameters.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 7, 2026, 02:32 PM
Security Audit — agent-trust-hub — detecting-entra-offensive-tools-in-graph-logs