AGENT SKILL

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

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

  1. Scope: the change and the parts of the system it touches.
  2. 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.
  3. Name the assets: data and capabilities worth protecting, and who must not reach them.
  4. Find threats per boundary crossing, using the categories in references/stride.md as a checklist, not as a form to fill.
  5. Decide per threat: mitigated (by what, where), accepted (by whom, why), or open. Prefer mitigations the architecture already uses.
  6. Turn mitigations into checks: each mitigation becomes a requirement with a test or a review point that security-review will verify.
  7. 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

04 同仓库更多 Skills