Threat modeling grounded in source code
Connect entry points, valuable assets and trust boundaries to concrete abuse scenarios. OpenAI’s workflow separates observed controls from assumptions.
27,227 · GitHubopenai/skillsEditorial selection · source reviewedAuthor: OpenAI Apache-2.0 Revision 49f948f
When to use it
Review a service architecture before launch
Analyze file-upload and background-job risks
Map isolation boundaries between tenants
What you’ll get
A Markdown report with threats, reasoned priorities, mitigations and open questions.
For example
Threat-model this document-upload service using its code and deployment design. Map trust boundaries and distinguish assumptions from confirmed vulnerabilities.
Original, requirements and limits
Original skill: OpenAI. Editorial description: kossolapov.com.
- Agent
- An agent that can read Agent Skills instructions; tool access depends on the host
- Tools
- Read access to the in-scope repository
- Deployment and authentication context from the service owner
- Limits
- A threat model is not a penetration test or proof that vulnerabilities exist.
- The workflow asks for confirmation of assumptions that affect scope or severity.
- Editorial selection based on source inspection, not a comparative performance certification.