domain-modeling
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user input and existing project documentation to build a domain glossary, which presents a surface for indirect prompt injection.
- Ingestion points: The agent reads user input regarding domain terms and the content of
GLOSSARY.mdas part of its workflow. - Boundary markers: Structural boundaries are provided by the
assets/glossary.template.mdfile using markdown headers and HTML comments. - Capability inventory: The skill is authorized to read project files, check source code for behavior claims, and perform file-write operations to update
GLOSSARY.md. - Sanitization: The
SKILL.mdinstructions include strict 'Entry rules' that explicitly forbid implementation details, requirements, and conversation narrative, acting as a constraint against the persistence of potentially malicious instructions embedded in domain terms.
Audit Metadata