data-pipeline-builder

Pass

Audited by Gen Agent Trust Hub on Aug 28, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests user-provided data sources, schemas, and transformation requirements to generate executable Python scripts, SQL files, and Airflow DAGs. A malicious user could provide requirements designed to inject harmful commands or exfiltrate data in the resulting pipeline implementation.
  • Ingestion points: Gathering requirements for data sources, destinations, and transformations in SKILL.md (Workflow step 1).
  • Boundary markers: Absent; there are no instructions to delimit user input or explicitly ignore embedded instructions within the provided requirements.
  • Capability inventory: The agent has access to Write and Bash tools, which it uses to create the implementation files and project structure defined in references/project-structure.md.
  • Sanitization: Absent; no explicit validation or sanitization of input requirements is performed prior to generating implementation code, although the workflow does include a step for the user to confirm the design (Workflow step 3).
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 28, 2026, 07:25 AM
Security Audit — agent-trust-hub — data-pipeline-builder