oracle-ai-architect
Pass
Audited by Gen Agent Trust Hub on Oct 13, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill provides architectures for ingesting data from OCI Object Storage and Oracle Autonomous Databases into GenAI pipelines. This creates an indirect prompt injection surface where untrusted data could influence agent behavior.
- Ingestion points: OCI Object Storage buckets (SKILL.md) and sales database tables (SKILL.md).
- Boundary markers: The provided reference code does not demonstrate the use of delimiters or 'ignore embedded instructions' warnings for ingested content.
- Capability inventory: The skill involves LLM inference via OCI Generative AI, SQL execution via Select AI, and data processing via Spark.
- Sanitization: No sanitization or content filtering is implemented in the provided architectural examples.
- [EXTERNAL_DOWNLOADS]: The skill references standard dependencies for Oracle AI development, including the official OCI SDK (
oci),pyspark,requests, and the Oracle ADK (oci-adk). These are well-known libraries associated with the targeted cloud ecosystem. - [COMMAND_EXECUTION]: The Terraform snippets include the use of
user_datato execute configuration scripts during instance provisioning. This is a standard infrastructure-as-code pattern for the skill's stated purpose of architecting OCI solutions.
Audit Metadata