threat-modeling
Identify assets, trust boundaries, likely threats and mitigations in a proposed design. Use when designing a new component, interface, data flow, integration or handling of sensitive data, or when the user asks for a threat model; use security-review for a completed code change.
- License 许可
- MIT
- 最近更新
- 2026-11-14
- Stars
- ★ 1
- Forks
- 0
01 前言 Front Matter
- name
- threat-modeling
- description
- Identify assets, trust boundaries, likely threats and mitigations in a proposed design. Use when designing a new component, interface, data flow, integration or handling of sensitive data, or when the user asks for a threat model; use security-review for a completed code change.
02 资源文件 3
- 参考文档examples/threat-model.md1.3 KB
- 参考文档references/stride.md1.9 KB
- 参考文档skill.yaml316 B
03 SKILL.md 正文
Threat modeling
When to use
In planning, for changes that alter how data or control crosses a trust boundary, and
before security-review when that review would otherwise have nothing to check against.
It works on the design; it does not need finished code.
Required context
A description of the change. Useful: the architecture and integrations context (existing components, boundaries and external systems) and the domain context (what data is sensitive).
Preconditions
The design is concrete enough to draw: which components, which data, which flows.
Procedure
- Scope: the change and the parts of the system it touches.
- Model: list the components, data stores, external actors and flows between them, and mark each trust boundary a flow crosses. A table is enough; a diagram helps.
- Name the assets: data and capabilities worth protecting, and who must not reach them.
- Find threats per boundary crossing, using the categories in references/stride.md as a checklist, not as a form to fill.
- Decide per threat: mitigated (by what, where), accepted (by whom, why), or open. Prefer mitigations the architecture already uses.
- Turn mitigations into checks: each mitigation becomes a requirement with a test or
a review point that
security-reviewwill verify. - Record the model in the shape of examples/threat-model.md, with unknowns listed.
Tools
None.
Rules
None beyond those that apply by scope. Acceptance of a risk is a human decision, never the agent's.
Verification
None is required: a threat model is a design artifact, checked later through the requirements it produced.
Evidence
A review artifact containing the model, the threats and their decisions.
Completion criteria
- Every flow that crosses a trust boundary was examined.
- Each threat is mitigated, accepted by a named human, or listed as open.
- Each mitigation has a way to be verified.
Failure modes
| Failure | Signal | Response |
|---|---|---|
| Checklist theater | Every category filled for every flow with generic text | Keep only threats specific to this design |
| Missing boundary | A flow to an external system has no threats listed | Re-draw the model; examine it |
| Agent accepts risk | A threat marked accepted with no human named | Mark it open; ask |
| Model without follow-up | Mitigations have no check | Add the requirement or review point |
References
- references/stride.md: threat categories and typical mitigations.
- examples/threat-model.md: a synthetic threat model.
04 同仓库更多 Skills
sample-skill
Does a sample task. Use when a sample is needed.
unexplained-contract
Does a sample task. Use when a sample is needed.
shell-block
Does a sample task. Use when a sample is needed.
project-identifier
Does a sample task. Use when a sample is needed.
oversized
Does a sample task. Use when a sample is needed.
no-activation
Does a sample task.