build-data-pipeline
Pass
Audited by Gen Agent Trust Hub on Oct 31, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill's primary function involves ingesting and processing untrusted external data from various producers, which introduces a surface for indirect prompt injection.
- Ingestion points: Raw data payloads from external producers are ingested and validated against ODCS v3.2.0 contracts at the perimeter before storage persistence as described in SKILL.md and references/producer-contracts-and-lakehouse.md.
- Boundary markers: The skill explicitly mandates the use of ODCS v3.2.0 contract verification, Write-Audit-Publish (WAP) on isolated snapshot branches, and schema invariance gates to isolate and validate untrusted input.
- Capability inventory: Across its files, the skill utilizes tools with file system access (write_file, edit_file, create_file), database operations (query_db, read_database), and testing capabilities (run_tests).
- Sanitization: The instructions include explicit guardrails to sanitize RAG vector inputs per OWASP ASI06 and mask or hash PII according to a data-classification.yaml policy.
Audit Metadata