Threat modeling

[anssiko]

@simoneonofri from the Security Interest Group volunteered to organize a Threat Modeling session for WebMCP.

Simone will provide further guidance in this meta issue to help the group improve its Security and Privacy Considerations for WebMCP with focus on the security part.

The group prefers smaller self-contained security issues for actionable things. Feel free to file new issues as appropriate.

[innotommy]

Hi @anssiko,
The next TMCG session is on April 7, 2026, where we can tackle WebMCP's security and privacy considerations together!
Here's what I'm thinking: we could start with a quick explainer covering the standard and the main concerns the group has already identified (prompt injection attacks, misrepresentation of intent, and privacy leakage through over-parameterization). Then we'd jump into an interactive threat modeling session using the same approach that worked well for DID, the STRIDE framework with that LEGO SERIOUS PLAY style (listen, think, share, reflect).
This way we can systematically identify threats and gather actionable feedback from the group.