event-credit-distribution
Pass
Audited by Gen Agent Trust Hub on Sep 18, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill reconciles data from Luma guest lists and Google Form responses to automate messaging through an XChat transport, which exposes the agent to untrusted external input.
- Ingestion points: The agent reads guest data from Luma guest-management URLs and registration answers, as well as participant responses from Google Form URLs or sheets (SKILL.md).
- Boundary markers: The instructions mandate a safety boundary by requiring explicit user authorization for recipients and message content before any sending occurs (SKILL.md).
- Capability inventory: The skill uses an 'XChat transport' for external communication and writes to a local 'delivery-checklist.md' tracking file (SKILL.md).
- Sanitization: The instructions include normalization of X handles for identity matching but do not explicitly define sanitization or escaping procedures for other ingested data before it is interpolated into message templates (SKILL.md).
Audit Metadata