investigating-replay
Pass
Audited by Gen Agent Trust Hub on Oct 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data which may contain embedded instructions designed to influence the agent's narrative or actions.\n
- Ingestion points: Exception messages, URL properties, and person metadata are ingested from the PostHog database via
posthog:execute-sqlandposthog:persons-retrieve(SKILL.md).\n - Boundary markers: The instructions lack explicit delimiters or warnings to treat ingested event data as untrusted content.\n
- Capability inventory: The skill possesses the capability to query databases (
posthog:execute-sql) and generate AI-driven summaries (posthog:vision-scanners-inline-scan).\n - Sanitization: There are no sanitization or filtering steps applied to the event properties or exception values before they are used to synthesize the final investigation report.
Audit Metadata