sdd4j-hexagonal
Pass
Audited by Gen Agent Trust Hub on Oct 7, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted architectural specifications and JavaDoc content from project source files, which can be manipulated to influence agent behavior.
- Ingestion points: The skill specifically reads and maps data from package-info.java files located within the project structure (SKILL.md).
- Boundary markers: The mapping logic relies on identifying headers such as ## Boundary, ## Requirements, and ## Entities, but it lacks instructions to sanitize or ignore arbitrary command instructions nested within these sections.
- Capability inventory: As an architecture adapter designed for composition with code-writing skills, the risk is that poisoned input in the specification could lead to unauthorized or malicious code modifications performed by the agent.
- Sanitization: The skill does not prescribe any validation, escaping, or sanitization of the values extracted from the source files before they are processed by the agent.
Audit Metadata