Skip to content
Book demo

Security

Threat model from PRD or RFC

Generate a STRIDE-style threat model from a product spec or technical design, then turn abuse cases into security requirements and review gates.

threat-modelingstrideappsecsecurity-reviewprdrfcprivacy

[ workflow / security ]

Threat model from PRD or RFC

Cosmos shifts security left by threat-modeling proposed changes before implementation. It reads specs, diagrams, data flows, and repo context; identifies assets, trust boundaries, abuse cases, and sensitive-data paths; then creates security acceptance criteria and routes high-risk work to AppSec.

07 nodes

06 edges

Trigger[trigger]
Spec enters security review

PRD / RFC / design doc

AI Agent step[extract]
Extract system model

Assets · actors · data flows

AI Agent step[stride]
Run STRIDE analysis

Threats by boundary + asset

AI Agent step[controls]
Define controls

Security requirements + tests

Decision

High-risk threats?

No
Output / Result[criteria]
Security acceptance criteria

Added to implementation tickets

YES
Human-in-the-loop[appsec]
AppSec approval

Required for Critical / High

Workflow prompt

Paste this into Augment to reproduce the workflow end-to-end.

Cosmos, whenever a PRD, RFC, or design document enters security review, create a threat model before implementation begins. Extract assets, actors, trust boundaries, data flows, authentication and authorization points, sensitive data, third-party integrations, and deployment surfaces. Apply STRIDE to identify spoofing, tampering, repudiation, information disclosure, denial-of-service, and elevation-of-privilege risks. For each credible threat, add abuse-case reproduction notes, required controls, logging and audit requirements, test cases, and the owning team. Route Critical or High risks to AppSec for approval; lower risks become acceptance criteria on the implementation tickets.