alterlab-genai-text-to-image
Pass
Audited by Gen Agent Trust Hub on Sep 21, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill instructions create a potential surface for indirect prompt injection by requiring the agent to ingest untrusted data from multiple sources.
- Ingestion points: The agent is directed in
SKILL.mdto "Read existing project files for context" (including scripts and mood boards) and to "Search the web for current Higgsfield models, prompt syntax updates, new features, and pricing changes". - Boundary markers: There are no explicit instructions or delimiters provided to help the agent distinguish between its core instructions and potentially adversarial content found in project files or web search results.
- Capability inventory: The skill leverages capabilities for reading local files, performing web searches, and writing structured markdown files (e.g.,
{project}-prompt-library.md) back to the environment. - Sanitization: The protocol lacks specific steps for sanitizing, validating, or filtering input from external project files or web content before it is used to generate new files or influence agent behavior.
Audit Metadata