model-integration

Pass

Audited by Gen Agent Trust Hub on Mar 4, 2026

Risk Level: SAFENO_CODEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it processes multiple data sources to generate its outputs. Ingestion points: The skill reads domain-alignment.json, domain-newconcepts.json, domain-concepts.json, requirements.json, and existing organizational models. Boundary markers: No markers or ignore instructions are present to prevent the agent from obeying instructions within the data. Capability inventory: The agent has the ability to write multiple JSON and Markdown reports. Sanitization: No sanitization steps are defined to filter input content.
  • [NO_CODE]: The skill is entirely documentation-based and does not provide any executable code scripts.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 4, 2026, 12:22 PM