| name | sota-threat-modeling |
| description | State-of-the-art threat modeling for both designing new systems and auditing existing ones. Use when designing a feature, service, integration, or architecture that touches untrusted input, new trust boundaries, sensitive data, or third-party dependencies (BUILD mode), and when reviewing, auditing, or pen-test-scoping an existing codebase to reconstruct its implicit threat model and find gaps (AUDIT mode). Not for code-level vulnerability review — use sota-code-security. Trigger keywords: threat model, STRIDE, LINDDUN, PASTA, attack tree, kill chain, data flow diagram, DFD, trust boundary, attack surface, abuse case, security design review, security architecture review, risk rating, DREAD, CVSS, security requirements, secure design, security audit, gap analysis, prompt injection, excessive agency, threat catalog, mitigations, residual risk. |
SOTA Threat Modeling
Purpose
Threat modeling answers Shostack's four questions with engineering rigor:
- What are we working on? (decompose: DFD, trust boundaries, assets, actors)
- What can go wrong? (enumerate: STRIDE/LINDDUN per element, catalogs, attack trees)
- What are we going to do about it? (treat: mitigate/accept/transfer/avoid, map to requirements and tests)
- Did we do a good job? (verify: abuse-case tests, residual risk review, re-model triggers)
This skill operationalizes those questions in two modes. Never produce a threat
model that is only prose — every threat must land as a tracked requirement, a
test, or an explicitly accepted risk with an owner.
BUILD Mode — Threat-Model-While-Designing
Run this workflow whenever designing anything that crosses a trust boundary.
Scale effort to risk: a 15-minute "four questions" pass for a small feature; a
full STRIDE-per-interaction model for a new service or auth flow.
Workflow
- Scope the delta. Model what is new or changed, not the whole system.
List new entry points, new data classes, new dependencies, new actors.
- Draw the DFD as text/mermaid (see
rules/02). Mark trust boundaries
explicitly. If you cannot draw a boundary, you do not understand the design
yet — stop and ask.
- Pick the methodology (see
rules/01): STRIDE-per-interaction by
default; add LINDDUN if personal data flows; attack trees for a single
high-value asset; four-questions-only for low-risk deltas.
- Enumerate threats crossing each boundary using the per-component
catalogs in
rules/03. Write each threat as: actor → action → asset →
impact. No vague entries ("hacking", "data breach").
- Rate and treat each threat (see
rules/04): likelihood × impact matrix,
then accept / mitigate / transfer / avoid. Every mitigation becomes a
security requirement with an ID.
- Emit artifacts (see
rules/05): threat model doc, security requirements
backlog entries, abuse cases as test stubs, and re-modeling triggers.
- Wire into delivery. Reference requirement IDs in the design doc, tickets,
and PR descriptions. A threat without a tracked artifact does not exist.
Continuous / incremental (agile, PR reviews)
- Threat model the story, not the sprint. Add a "Security notes" section to
design docs and PR descriptions for any change matching a re-model trigger:
new dependency, new endpoint/route/queue/cron/webhook, new trust boundary,
new data class, auth/authz change, file/deserialization handling.
- In PR review, run a micro-STRIDE on the diff only: what new input enters?
whose privilege executes it? what does it write or call? Takes 5 minutes;
catches the majority of design-level regressions.
AUDIT Mode — Reconstructing a Threat Model from Code
Use when handed an existing system with no (trustworthy) threat model. Goal:
rebuild the implicit model from artifacts, then diff intended vs. actual
controls. Full procedure in rules/06.
Workflow
- Inventory entry points from code (see
rules/02 §extraction): routes,
queue consumers, cron jobs, webhooks, third-party callbacks, CLI/admin
tools, file uploads, IaC-exposed ports.
- Reconstruct the DFD from the inventory: processes, stores, external
entities, flows; infer trust boundaries from network topology, authn
checkpoints, and IAM policies.
- Identify assets and actors from schemas, secrets handling, and config.
- Run the catalogs (
rules/03) against each component; for every catalog
item record: control present / absent / partial, with file:line evidence.
- Gap analysis: rank absent/partial controls by exploitability ×
blast radius; distinguish "missing control" from "missing defense-in-depth".
- Report findings in the standard format below.
Severity conventions
| Severity | Definition |
|---|
| Critical | Remotely exploitable now, by an unauthenticated or low-priv actor, leading to full compromise of a key asset (RCE, auth bypass, mass data exfil). Fix before anything else ships. |
| High | Exploitable with realistic preconditions (one valid account, one misconfig, MitM position) compromising a key asset; or a Critical with a single weak mitigating layer. Fix this sprint. |
| Medium | Requires chaining, elevated access, or unusual conditions; or impacts a secondary asset; or defense-in-depth gap on a Critical path. Schedule. |
| Low | Hardening, hygiene, info disclosure of low-value data, theoretical with strong existing controls. Backlog. |
Severity = exploitability × impact in this deployment context — never copy
a CVE/CVSS base score without environmental adjustment (see rules/04).
Finding format (every finding, no exceptions)
[SEV] TITLE (component, STRIDE/LINDDUN class)
Location: path/to/file.py:123 (and IaC/config refs)
Threat: <actor> can <action> via <vector> because <missing/weak control>,
impacting <asset> (<C/I/A/privacy impact>).
Evidence: code excerpt or config line proving the gap.
Recommendation: specific control + where it goes; map to requirement ID.
Residual risk if accepted: one sentence.
Rules Index
| File | Read this when... |
|---|
rules/01-methodologies.md | Choosing between STRIDE, LINDDUN, PASTA, attack trees, kill chains; deciding lightweight vs. heavyweight; setting up continuous/PR-level threat modeling. |
rules/02-decomposition.md | Drawing DFDs in mermaid, defining trust boundaries, listing entry points/assets/actors/privilege levels; extracting all of these from an existing codebase. |
rules/03-threat-catalogs.md | Enumerating threats for a specific component: web frontend, API, database, message queue, file storage, CI/CD, mobile, LLM agent/tool-use, cloud/IAM. |
rules/04-risk-rating-treatment.md | Rating threats (DREAD pitfalls, CVSS usage, L×I matrices), choosing accept/mitigate/transfer/avoid, mapping mitigations to requirements and tests, documenting residual risk. |
rules/05-outputs-operationalization.md | Writing the threat model document, building the security requirements backlog, turning abuse cases into tests, keeping the model alive (re-model triggers). |
rules/06-audit-reconstruction.md | Auditing an existing system: reconstructing the model from code, control-presence matrix, gap analysis, severity calibration, reporting. |
Load only the files you need; rules/02 + rules/03 cover 80% of day-to-day work.
Top-10 Non-Negotiables
- No model without a diagram. Every threat model includes a DFD (mermaid
or ASCII) with explicit trust boundaries. Prose-only models hide boundary
confusion.
- Threats are sentences, not nouns. Actor → action → asset → impact.
"SQL injection" is a vector; "anonymous user exfiltrates the orders table
via unparameterized search query" is a threat.
- Every entry point gets enumerated — including queues, cron, webhooks,
callbacks, admin tooling, and CI/CD. HTTP routes are never the whole attack
surface.
- Trust boundary crossings drive enumeration. Apply STRIDE per
interaction at each crossing; data inside one boundary at one privilege
level rarely needs the full treatment.
- Personal data ⇒ LINDDUN pass. STRIDE does not cover linkability,
identifiability, or non-compliance; run a privacy pass whenever PII flows
or is stored.
- Rate with likelihood × impact in context. Never ship raw DREAD scores
or unadjusted CVSS base scores as priorities.
- Every threat gets a disposition. Mitigate (→ requirement ID + test),
accept (→ named owner + expiry date), transfer, or avoid. "Noted" is not a
disposition.
- Mitigations become tests. Each mitigated threat yields at least one
abuse-case test (unit, integration, or rule-based check) that fails if the
control regresses.
- LLM/agent components are first-class attack surface. Model prompt
injection, tool-call abuse, excessive agency, and data exfil via outputs
for any system invoking an LLM with tools or retrieved content.
- Models expire. Define re-model triggers (new dependency, new trust
boundary, new data class, auth change) in the document itself; an undated,
trigger-less threat model is treated as absent in audits.