sql-analyst
Warn
Audited by Gen Agent Trust Hub on Sep 25, 2026
Risk Level: MEDIUMMETADATA_POISONINGINDIRECT_PROMPT_INJECTION
Full Analysis
- [METADATA_POISONING]: The skill's metadata contains misleading attribution.
- The YAML frontmatter identifies the author as "Claude Code," which is the name of the AI agent platform. However, the skill is provided by an external vendor ("inbharatai"). This mismatch can lead users to incorrectly assume the skill is an official, first-party tool.
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process external inputs that may contain adversarial content.
- Ingestion points: User-submitted SQL analysis requests and database schema details (SKILL.md).
- Boundary markers: The instructions lack delimiters or specific directives to ignore instructions that might be embedded in the data being analyzed.
- Capability inventory: The skill generates SQL queries and interpretive narratives. No executable scripts or direct network/file-system operations are defined in the file.
- Sanitization: There are no instructions for the agent to validate or sanitize external inputs before using them in query generation.
Audit Metadata