agentic-security:threat-model-first

Threat modeling before writing code involves identifying security risks at trust boundaries using the STRIDE framework. You pause implementation to document spoofing, tampering, and elevation of privilege risks upfront, generating a TM.md artifact to guide secure development.

STRIDE coverage is a structured threat modeling process evaluating Spoofing, Tampering, Repudiation, Information disclosure, Denial of service, and Elevation of privilege. You need it before editing code that handles authentication, authorization, secrets, or external API calls to prevent design flaws.

You document security risks by creating a session, walking through STRIDE categories for each touch-point, and writing the results to a TM.md file. This artifact captures the identified threats and proposes defensive measures with code comment annotations.

Yes, you can map STRIDE threat modeling results directly into code comments. As you implement defensive measures against identified threats, you reference each defense in the code comments to show the mapping to specific STRIDE rows.

The best way to front-load security design is to pause before editing code, generate a threat model session, and perform a STRIDE walkthrough for the specific construct. This pre-write risk assessment ensures the implementation is informed by documented mitigations.

No dedicated environment is needed beyond your standard development setup and MCP scratchpad access. The threat modeling process runs within your conversation, generating a session ID and storing the TM.md documentation artifact directly through the MCP integration.