understanding-billing-usage
Pass
Audited by Gen Agent Trust Hub on Sep 4, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill mentions dispatching tools via
mcp__posthog__exec. However, this is part of the standard Model Context Protocol (MCP) dispatching mechanism for this platform and does not execute arbitrary shell commands or untrusted external code. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted customer data such as event names, property values, and logs. It explicitly mitigates this risk by instructing the agent to treat these as evidence only and to not follow instructions embedded within that data.
- [SAFE]: The SQL query provided in Step 5 is a template for querying the customer's own events database. It includes a whitelist of excluded billable events to ensure accurate reporting and does not allow for arbitrary data exfiltration or deletion.
Audit Metadata