Security

Create the threat model. Argument: $ARGUMENTS

Context Gathering

Before interviewing, understand the system's security surface:

1. Check for existing docs:

• Read docs/THREAT_MODEL.md — if revising, don't start from scratch
• Read docs/ARCHITECTURE.md for system structure and trust boundaries
• Read CLAUDE.md, README.md

1. Explore the codebase for security-relevant patterns:

• Look for auth middleware, JWT handling, session management
• Check for input validation patterns (or lack thereof)
• Look for database queries (raw SQL vs ORM/parameterized)
• Check for secrets management (env vars, config files, hardcoded values)
• Check for exposed endpoints (route definitions, API handlers)
• Read docker-compose.yml or deployment config for service exposure
• Summarize what you found to the user — "Here's what I see from a security perspective: ..."

Interview

Use AskUserQuestion to understand the security landscape. Adapt based on what the codebase reveals.

1. Trust boundaries — What is trusted? What is untrusted? Where are the boundaries?
2. Authentication — How do users/agents/services prove identity? What mechanisms exist today?
3. Authorization — Who can do what? How are permissions modeled?
4. Sensitive data — What data is sensitive? Credentials, PII, financial? Where does it live? How is it protected at rest and in transit?
5. Attack surface — What is exposed to the internet? To local users? To other services? What inputs does the system accept?
6. Threat actors — Who would attack this? Script kiddies, insiders, nation states? What are their capabilities?
7. Compliance — Any regulatory requirements? SOC2, GDPR, HIPAA, PCI?
8. Existing security measures — What's already in place? What's missing?

For inherited projects: "I see you're using X for auth — is that intentional or legacy? I notice Y has no input validation — is that a known gap?"

Write

Write to docs/THREAT_MODEL.md:

# Threat Model

**Version:** [version]
**Date:** [date]

## Trust Assumptions

[What is the trust model? What is trusted, what is untrusted? Include a Mermaid diagram of the trust hierarchy.]

## Security Properties

[Non-negotiable security invariants. Things that must always hold.]

- [Property 1]
- [Property 2]

## Attack Surface

| Surface | Exposure | Controls |
|---------|----------|----------|
| [surface] | [who can reach it] | [what protects it] |

## Threats

### [Threat Category 1: e.g., Agent Impersonation]

| # | Attack | Impact | Likelihood | Defense |
|---|--------|--------|------------|---------|
| 1 | [attack description] | [what happens] | [Low/Med/High] | [how it's prevented] |

### [Threat Category 2]
...

## Sensitive Data Inventory

| Data | Classification | At Rest | In Transit | Access Control |
|------|---------------|---------|------------|----------------|
| [data type] | [level] | [protection] | [protection] | [who can access] |

## Security Controls

### Implemented
- [Control 1 — what it protects against]

### Required (not yet implemented)
- [Control 1 — what it would protect against, priority]

## Compliance Requirements

[If applicable. Regulatory frameworks, what they require, current status.]

After Writing

1. Present the document to the user for review. Iterate until they're satisfied.
2. Suggest next steps based on what exists:

• No architecture doc? → "Define system structure with /architecture"
• No TDD? → "Define testing and dev workflow with /tdd"
• Ready to build? → "Create feature specs with /spec , then /roadmap"

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: 0xrafasec
• Source: 0xrafasec/ai-workflow
• License: MIT

Install and usage instructions live in the source repository linked above.