turning-engineering-analytics-into-insights

Pass

Audited by Gen Agent Trust Hub on Jul 21, 2026

Risk Level: SAFE
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses the execute-sql tool to run HogQL queries against GitHub metadata. These queries are primarily used to test and validate data before creating permanent insights.
  • [DATA_EXFILTRATION]: The skill accesses GitHub PR and CI/CD data stored in PostHog warehouse tables (github_pull_requests, github_workflow_runs). The analysis confirmed that this data is processed internally using PostHog's insight-create and dashboard-create tools and is not sent to external or untrusted destinations.
  • [PROMPT_INJECTION]: An analysis of the indirect prompt injection surface was performed. The skill processes data from GitHub tables that could contain user-controlled strings (e.g., PR titles). Because this data is used for structured visualization (charts and graphs) rather than as a source of instructions for the agent, the risk of instruction override is minimal.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 21, 2026, 08:12 PM
Security Audit — agent-trust-hub — turning-engineering-analytics-into-insights