dbt-model-generator
Fail
Audited by Gen Agent Trust Hub on Feb 16, 2026
Risk Level: HIGHPROMPT_INJECTIONNO_CODE
Full Analysis
- [Indirect Prompt Injection] (HIGH): The skill is designed to process external content (dbt models and data pipeline requests) while possessing high-privilege capabilities including
Bash,Write, andEdittool access. This creates a significant vulnerability surface where instructions hidden in untrusted data could be executed with the agent's permissions. - Ingestion points: User requests containing dbt model definitions or patterns.
- Boundary markers: Absent; no instructions are provided to the agent to treat external data as untrusted or to ignore embedded instructions.
- Capability inventory: Access to
Bash(command execution),Write/Edit(file system modification), andRead/Grep(data exposure). - Sanitization: Absent; no logic is provided to validate or sanitize inputs before they are processed by high-privilege tools.
- [No Code] (LOW): This is a metadata-only skill consisting exclusively of the
SKILL.mdconfiguration file. While no malicious scripts are included in the package itself, the requested tool permissions are excessive for a skill providing only instructional guidance.
Recommendations
- AI detected serious security threats
Audit Metadata