pubnub-illuminate
Pass
Audited by Gen Agent Trust Hub on Sep 28, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill architecture is designed to ingest and act upon untrusted data from PubNub message streams, which represents a vulnerability surface for indirect prompt injection.\n
- Ingestion points: Message bodies ($.message.body), metadata, user IDs, and channel names are extracted from incoming PubNub messages to be evaluated by Illuminate Business Objects and Decisions as documented in SKILL.md and references/business-objects.md.\n
- Boundary markers: The provided instructions do not specify the use of boundary markers or delimiters to separate untrusted message content from the system's logic or prompt context.\n
- Capability inventory: The skill possesses significant side-effect capabilities across multiple scripts, including triggering external webhooks via WEBHOOK_EXECUTION, publishing new messages via PUBNUB_PUBLISH, and modifying user/channel metadata within the PubNub App Context (references/decisions-4-step-workflow.md).\n
- Sanitization: There are no explicit requirements or instructions for sanitizing or validating the ingested message data before it is utilized in Decision conditions or interpolated into Action templates.
Audit Metadata