realistic-prompt-generation
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill defines a controlled process for generating natural language prompt variants based on specific architectural constraints. It focuses on maintaining intent and context while explicitly prohibiting the use of target brands or proprietary data.
- [INDIRECT_PROMPT_INJECTION]: The skill processes external data structures (
blind_design_brief.jsonandprompt_architecture.json). It provides explicit instructions for the agent to detect and reject unauthorized or sensitive fields (e.g., target brands, slogans, or campaign terms) within these inputs. The inclusion of these boundary checks and the absence of high-risk capabilities like network access or file-system writing mitigate the risk of indirect prompt injection.
Audit Metadata