engineering-gameplay-analytics
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill implements an event tracking system that ingests untrusted data from external game clients, creating a surface for indirect prompt injection.
- Ingestion points: The track method in boilerplate/analytics-client.ts accepts an arbitrary properties object from external callers, which is stored in the database.
- Boundary markers: There are no instructions or delimiters provided to ensure that an agent analyzing this data treats the properties as untrusted content.
- Capability inventory: The skill includes database write operations and complex read queries that could be influenced if an agent is tricked by malicious instructions within the event stream.
- Sanitization: The event properties are passed directly to storage as a JSONB payload without validation or filtering.
- [DATA_EXFILTRATION]: The query templates in templates/retention-dashboard.md exhibit a potential data exposure risk through unsafe SQL patterns.
- Evidence: The retentionByDay, computeAndStoreCohort, and dauMau functions use sql.raw to interpolate interval values (e.g., dayOffset and days) into SQL queries. If these parameters are derived from untrusted user input without strict validation, it could facilitate SQL injection attacks that expose sensitive database contents.
Audit Metadata