oracle-db-skills
Pass
Audited by Gen Agent Trust Hub on Sep 4, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The INSTALL.ps1 script fetches the skill repository from GitHub. This is a standard installation method for repository-based skills.
- [INDIRECT_PROMPT_INJECTION]: The skill provides instructions for agents to process database schema and content, creating an attack surface for indirect prompt injection from the target database. Ingestion points: Data enters the agent's context through SQL query results as described in agent/schema-discovery.md and agent/nl-to-sql-patterns.md. Boundary markers: The skill includes robust guidance in agent/intent-disambiguation.md and agent/safe-dml-patterns.md instructing the agent to verify intent and row counts before acting on data. Capability inventory: The guides describe powerful capabilities such as EXECUTE IMMEDIATE and DBMS_SQL (in sql-dev/dynamic-sql.md) and OS-level interactions via UTL_FILE. Sanitization: The skill provides extensive documentation on preventing SQL injection using DBMS_ASSERT and bind variables in sql-dev/sql-injection-avoidance.md.
Audit Metadata