schema-authoring
Pass
Audited by Gen Agent Trust Hub on Aug 28, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill provides instructional content and templates for Vespa search engine schemas. All analyzed files are restricted to documentation and static configuration examples.
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to analyze and evolve user-provided schema files, creating a standard surface for indirect prompt injection via data ingestion. However, this is a routine aspect of developer-centric tools.
- Ingestion points: The agent reads and processes
.sdschema files located within the project'sschemas/directory. - Boundary markers: The skill does not explicitly define markers to separate user-provided file content from agent instructions, relying on the structured nature of schema files.
- Capability inventory: The skill facilitates file read/write operations for development purposes but does not trigger shell execution, network exfiltration, or privileged operations.
- Sanitization: No specific content sanitization or filtering logic is included in the instructions.
- [NO_CODE]: The skill does not distribute or execute scripts (Python, Node.js, or Shell). All code snippets are intended as templates for the user's Vespa application configuration.
Audit Metadata