agent-usage

Warn

Audited by Gen Agent Trust Hub on Jul 3, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONREMOTE_CODE_EXECUTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses the Bash tool to run shell commands for monitoring agent usage.
  • [DYNAMIC_EXECUTION]: The execution path for the script agent-usage-monitor.py is constructed dynamically using the AGENTIC_ENGINEERING_HOME environment variable. This allows for the execution of code from paths determined at runtime rather than static, verified locations.
  • [UNVERIFIABLE_DEPENDENCIES_AND_REMOTE_CODE_EXECUTION]: The skill relies on a script (agent-usage-monitor.py) located in a specific repository path that is not part of the skill distribution. The security and integrity of this script cannot be verified from the skill content alone.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes external data in the form of local session ledgers and provider quota APIs. Ingestion points: Local token ledgers and provider API responses (specified in SKILL.md). Boundary markers: No specific boundary markers or 'ignore' instructions are present to separate untrusted usage data from agent instructions. Capability inventory: The skill has access to the Bash tool across its scripts to execute system commands. Sanitization: There is no evidence of sanitization or validation of the usage data before it is reported to the user.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Jul 3, 2026, 05:44 AM
Security Audit — agent-trust-hub — agent-usage