data-context-extractor

Pass

Audited by Gen Agent Trust Hub on Sep 14, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • Indirect Prompt Injection Surface: The skill is designed to ingest tribal knowledge from analysts and metadata from database schemas to generate new agent instructions. This workflow creates a surface for indirect prompt injection, where malicious text in source data could be interpreted as instructions. (1) Ingestion points: Analyst responses and database metadata retrieved through warehouse tools. (2) Boundary markers: The templates currently lack explicit delimiters or instructions for the agent to disregard commands embedded in source data. (3) Capability inventory: The skill can generate SQL, create documentation, and package files via a Python utility. (4) Sanitization: There is no defined validation or escaping for the data collected before it is included in the generated skill files.
  • Local File Packaging Utility: The included scripts/package_data_skill.py utility performs standard filesystem reads and writes to create ZIP archives. This functionality is consistent with its role as a packaging tool for the generated data context assets.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 14, 2026, 05:10 PM
Security Audit — agent-trust-hub — data-context-extractor