Threat Studio

A live text-to-diagram editor for threat modelling. Paste a data-flow model in plain text on the left; get a rendered trust-boundary diagram and multi-framework analysis on the right - STRIDE, LINDDUN, CIA, DREAD risk scoring, an auto-scaffolded attack tree, a Cyber Kill Chain view, and control mapping - all from the same source. Export the diagram as PNG and the threat tables as Markdown/CSV. No build step, no dependencies, nothing leaves the browser.

It speaks the same threatmodel DSL as the threat-modelling skill, so a model you draft with the skill renders here unchanged.

Run it

• Locally (source): open threat-studio/index.html in any modern browser (double-click, or xdg-open threat-studio/index.html).
• Use the included claude skill to generate new diagrams from your codebase.

Project layout

The source lives in threat-studio/

threat-studio/
├── index.html   # markup
├── styles.css   # styles (light/dark tokens)
├── app.js       # parser, renderer, STRIDE analysis, PNG export
├── build.py     # inlines the three into ../threat-studio.html
└── CLAUDE.md    # working notes for Claude Code (build rule, architecture)
threat-studio.html   # generated bundle - do not edit by hand

.claude/skills/threat-modelling/
├── SKILL.md                          # skill workflow (Four Questions + STRIDE)
├── references/
│   ├── diagram-syntax.md             # threatmodel DSL + common mistakes
│   └── stride-per-element.md         # STRIDE definitions + mitigations
└── templates/
    ├── threat-model-template.md      # report structure (EN)
    └── threat-model-template.nl.md   # report structure (NL)

Use it

1. Pick a starting point from the Example dropdown, or type your own.
2. Edit the model source - the diagram, validation status, and Analysis tab update live. Invalid lines are flagged in the gutter and listed under the editor.
3. Switch to the Analysis tab and pick one or more frameworks from the chip bar (combine as many as you like): STRIDE, LINDDUN, CIA, DREAD, attack tree, kill chain, control mapping. Each renders from the same DFD.
4. Click PNG to export the current diagram (2× resolution, current theme), or -Markdown / -CSV on the Analysis tab to export the threat tables.
5. The Docs tab documents the full syntax and every lens.

Language

threatmodel "Title"            # header (alias: stride); title optional

external: User                 # external entity  (rectangle)   - id optional
process 1.1: Web Server        # process          (circle)
datastore D1: User DB          # data store        (open store) - R* if log/audit/journal

User -> 1.1 : "HTTPS Request"  # directed flow  (label required)
1.1 <-> D1 : Read/Write        # bidirectional flow

boundary "Internet" { User }   # trust boundary (renders left-to-right)
boundary "DMZ" { 1.1 }

Flows that cross a trust boundary are drawn in red and flagged as priority.

Rules enforced live

• every flow needs a label
• no store→store or external→external flows
• every flow endpoint must be declared
• an element belongs to at most one boundary
• no duplicate ids

Analysis frameworks

The Analysis tab carries a framework picker - toggle any combination; each lens renders from the same threatmodel DFD (the DSL is unchanged). Adding a lens is one entry in the FRAMEWORKS registry in app.js - the pluggable seam.

Lens	Kind	What it shows
STRIDE	per-element matrix	Security threats per element (default).
LINDDUN	per-element matrix	Privacy threats - Linkability, Identifiability, Non-repudiation, Detectability, Disclosure, Unawareness, Non-compliance. Unawareness maps to the data subject; the rest to processes/stores/flows.
CIA	per-element matrix	Confidentiality / Integrity / Availability per data-handling asset.
DREAD	interactive scoring	Editable Damage/Reproducibility/Exploitability/Affected-users/Discoverability (each 1–3), pre-filled from a heuristic, ranked by total risk. Session-only; resets on Example change.
Attack tree	auto-scaffold	Attacker-goal tree built from high-value assets + boundary-crossing flows - a starting point to refine.
Kill chain	phase view	DFD elements mapped to Cyber Kill Chain phases (illustrative).
Controls	reference tables	Each category → OWASP ASVS chapters + MITRE ATT&CK techniques; LINDDUN → privacy controls (GDPR/ISO 27701). Indicative.

Export: on the Analysis tab, -Markdown and -CSV export the active lenses' threat tables (the Markdown feeds the companion skill's report). The PNG button still exports the diagram.

Files

• threat-studio/ - editable source (index.html, styles.css, app.js) plus build.py.
• .claude/skills/threat-modelling/ - the companion Claude skill that produces full timestamped threat-model reports (EN/NL) using the same DSL.