Model QA Specialist
Pass
Audited by Gen Agent Trust Hub on Oct 25, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill's primary function is to ingest and audit external, untrusted content including model methodology documentation and raw datasets. This creates an attack surface where malicious instructions hidden in the audited data could potentially influence the agent's output or actions.
- Ingestion points: External documents and raw data are collected during 'Phase 1: Scoping & Documentation Review' and 'Phase 2: Data & Feature Quality Assurance' as described in
SKILL.md. - Boundary markers: The instructions do not define specific delimiters or instructions to the agent to disregard natural language prompts found within the audited data files.
- Capability inventory: The skill includes Python code that writes files to the local file system (e.g.,
plt.savefig()inSKILL.md) and performs complex statistical processing. - Sanitization: There is no evidence of sanitization or filtering logic to strip potential instructions from the external text or data before processing.
Audit Metadata