mdpi-hypernetwork-archetype
Pass
Audited by Gen Agent Trust Hub on Jul 5, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill processes untrusted user-generated content (posts and comments) to assign behavioral archetypes, creating an indirect prompt injection surface.
- Ingestion points: Text content ingested into 'content_df' or 'user_df' for analysis in Step 3 (Sentiment) and Step 4 (Toxicity) as described in 'SKILL.md'.
- Boundary markers: No specific boundary markers or 'ignore' instructions for the analyzed content were identified in the methodology or code snippets.
- Capability inventory: The skill records analysis results to a local file path 'docs/analysis/14-mdpi-hypernetwork-archetype.md'. No high-risk capabilities like subprocess execution or network egress using the raw input data were found.
- Sanitization: The methodology uses NLP scorers (VADER and Detoxify) to convert text to numeric values. While this abstracts the content, the skill lacks explicit sanitization to prevent the raw text from influencing the agent if it is ever evaluated directly.
Audit Metadata