sense-bases
Pass
Audited by Gen Agent Trust Hub on Sep 20, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill defines a framework for the agent to process untrusted
.basefiles (YAML) and generate executable SQL queries, creating a surface for indirect prompt injection. - Ingestion points: External
.basefiles and view configurations described in SKILL.md. - Boundary markers: No explicit delimiters for untrusted content are defined, but the structured translation process provides inherent logic boundaries.
- Capability inventory: The resulting SQL queries are executed against a local database containing file metadata and frontmatter; the skill refers to the 'sense' tool for execution.
- Sanitization: The skill advocates for secure practices, specifically recommending the use of bound parameters (
?) for dynamic inputs and schema validation usingpragma_table_infoto ensure only valid columns are queried.
Audit Metadata