data-pipeline-manager
Pass
Audited by Gen Agent Trust Hub on May 14, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill follows machine learning research best practices for data documentation and validation, with no evidence of malicious patterns or safety bypasses.\n- [SAFE]: Tool usage (Bash, WebFetch, Read, Write) is strictly aligned with the documented tasks of auditing local scripts, fetching benchmark data, and maintaining pipeline records.\n- [SAFE]: Indirect Prompt Injection Surface: The skill ingests untrustworthy data from local project files (e.g., data_utils.py) and external web sources. Ingestion points: Project root scripts and WebFetch tool outputs. Boundary markers: Not specified. Capability inventory: Bash, Write, and Edit tools. Sanitization: No explicit steps defined. The surface is appropriate for the skill's intended purpose and does not present elevated risk.\n- [SAFE]: All external tools mentioned (e.g., bigcode-evaluation-harness, lm-contamination-checker) are standard utilities in the machine learning community.
Audit Metadata