onyx-ai
Pass
Audited by Gen Agent Trust Hub on Nov 1, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill follows standard deployment practices for the Onyx AI platform. It uses official source code repositories and standard container orchestration tools without any suspicious modifications or hidden commands.
- [INDIRECT_PROMPT_INJECTION]: The skill describes a system that ingests data from external, potentially untrusted sources including web scraping, Slack, and Confluence. This creates an attack surface for indirect prompt injection, where an attacker could place malicious instructions in documents indexed by the system.
- Ingestion points: Data connectors listed in
SKILL.md(Confluence, Slack, Google Drive, GitHub, Web scraping). - Boundary markers: None explicitly defined in the deployment instructions.
- Capability inventory: LLM-based query processing over ingested documents.
- Sanitization: Not addressed in the basic setup guide.
Audit Metadata