agentic-security:threat-model-first

What problem does it solve?

Activates before you add or edit code that crosses a security trust boundary. The cheapest place to fix a security bug is the moment before you write it. This skill front-loads the threat modeling so the implementation that follows is informed.

Core Features & Use Cases

• Pause before the Edit. Don't write the code yet. Tell the user you're going to threat-model first.
• Generate a session id if one doesn't exist for this conversation.
• Walk STRIDE per touch-point. For the specific construct the user is about to introduce, work through: S (Spoofing), T (Tampering), R (Repudiation), I (Information disclosure), D (Denial of service), E (Elevation of privilege) etc., and document one sentence per category.
• Write the result to the scratchpad via MCP and store the threat model artifact.
• Propose defensive measures and annotate code comments to show mapping to STRIDE rows.
• Commit-as-you-go: reference each defense in code comments as you implement it.

Quick Start

Before you write code that touches a security trust boundary, run threat-modeling to generate a TM.md that captures STRIDE questions and mitigations.