skills.rest
akoita
akoitaCOMMUNITY
Explore all skills by this author →
𝕏@aboobakarGitHub

security-threat-model

Builds repository-grounded threat models with trust boundaries, ranked abuse paths, and mitigations.

Updated Jul 9, 2026
Search Tags:#security#threat modeling#stride#risk analysis#attack surface#pytm#trust boundaries
One-click install
npx skills add https://github.com/akoita/agent-toolkit --skill security-threat-model-akoita
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: security-threat-model
Source: https://github.com/akoita/agent-toolkit/tree/main/plugins/claude/security/skills/security-threat-model
Command: npx skills add https://github.com/akoita/agent-toolkit --skill security-threat-model-akoita

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? Threat models often describe a system nobody actually built, with severity rankings nobody can justify. This Skill extracts the real system model from the checked-in code, derives trust boundaries and entry points with file:line evidence, and produces a ranked, defensible threat model instead of a plausible guess. ## Core Features & Use Cases - Evidence-based system modeling: Extracts components, data stores, entry points, and trust boundaries directly from the repository, citing file:line for every claim and separating runtime from build tooling. - Calibrated threat enumeration and ranking: Defines attacker capabilities and non-capabilities, writes threats as end-to-end abuse paths, and ranks them by likelihood and impact with explicit assumptions. - Method and tooling support: Applies STRIDE, LINDDUN GO, or attack trees, generates model-as-code with pytm or threagile, and includes dedicated guidance for systems containing AI agents (MAESTRO layers, lethal trifecta). - Use Case: Point it at a web service repository before launch; it maps unauthenticated entry points, walks abuse paths to sensitive assets, pauses to confirm deployment assumptions with you, then delivers a ranked report separating existing mitigations from recommended ones. ## Quick Start Ask the agent to threat model this repository, identifying trust boundaries and ranking the top abuse paths with evidence from the code.

Frequently Asked Questions about security-threat-model

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I create a threat model from an existing codebase?▼

Extract the system model from the repository first: entry points, data stores, external integrations, and authentication code, each cited with file:line. Then derive trust boundaries, calibrate attacker capabilities, enumerate abuse paths with STRIDE, and rank by likelihood and impact.

STRIDE vs LINDDUN vs attack trees: which threat modeling method should I use?▼

Use STRIDE per component and data flow as the default for design review. Add LINDDUN GO when personal data is central, and use an attack tree when one high-value asset dominates the risk. Avoid PASTA for small teams.

How do I threat model a system that contains AI agents or LLMs?▼

Model the agent by what its tools can reach, not what its prompt says. Check for the lethal trifecta of private data access, untrusted content, and external communication, and walk the seven MAESTRO layers alongside STRIDE for conventional components.

Can I generate a threat model as code with pytm or threagile?▼

Yes. pytm v1.4.0 is the default: a Python script that outputs JSON, DFD diagrams, and reports, requiring Python 3.11 and graphviz. threagile takes one YAML model and produces PDF, Excel, and JSON, though its last release was v0.9.1 in July 2024.

When should I not use threat modeling?▼

Do not use threat modeling to find concrete vulnerabilities in code; use a repository security audit or diff review instead. It also does not replace scanners or the detailed LLM, agent, and MCP control checklists, which belong to dedicated skills.

How reliable is LLM-assisted threat modeling?▼

One controlled study reports 63-72% accuracy and 52-62% F1 for LLM-assisted STRIDE modeling, with failures in perspective, second-order threats, and enumeration. LLMs are reliable for breadth-first enumeration and model authoring, but humans should rank risk and decide priorities.

Related Skills

Explore More Software Engineering Skills→
dougdliudougdliu

security-threat-model

Identify components, data stores, and trust boundaries to generate a Markdown threat model.

Community
Advanced
SwitchbackTechSwitchbackTech
228

Codex-a11y-audit

Audit UI diffs for accessibility regressions and propose minimal fixes.

Official
Intermediate
mostafasudomostafasudo
3

cso

Audit infrastructure security across secrets, dependencies, CI/CD, and AI systems.

Community
Advanced
habrahgithubhabrahgithub

penetration-testing

Perform authorized penetration testing on applications, APIs, and network services.

Community
Advanced