跳至正文
OpenAgentSkill
浏览已开放 · 目录记录日期 2026-11-21 · 自动更新和账户服务暂时关闭。
金融

threat-model-authoring

Draft phase 3.5 of a threat model from the orientation brief, surface analysis, and maintainer answers. USE WHEN writing threat-model.md to the canonical §1.1–§1.19 structure. Combines concise prose with the §1.7 trust table and contract matrix, §1.8 output statements, §1.17 disposition table, §1.1 triager quick-start, and any prior-policy back-map. Tags every non-trivial claim as documented, maintainer, or inferred; maps inferred claims to §1.18; and maintains confidence counts. DO NOT USE FOR: sidecar generation, backtesting, or triage.

alpha-omega-security · 已有静态检查记录

安装与使用

运行命令前先阅读来源与依赖说明。复制命令不会自动安装或执行。

npx skills add alpha-omega-security/threat-model --skill threat-model-authoring

技能说明


name: threat-model-authoring description: >- Draft phase 3.5 of a threat model from the orientation brief, surface analysis, and maintainer answers. USE WHEN writing threat-model.md to the canonical §1.1–§1.19 structure. Combines concise prose with the §1.7 trust table and contract matrix, §1.8 output statements, §1.17 disposition table, §1.1 triager quick-start, and any prior-policy back-map. Tags every non-trivial claim as documented, maintainer, or inferred; maps inferred claims to §1.18; and maintains confidence counts. DO NOT USE FOR: sidecar generation, backtesting, or triage. argument-hint: ''

Threat Model — Authoring (draft the document)

Phase 3.5. Write the deliverable to the section structure in output-structure.md. Read principles.md first — the style bar is "describe the project as it is, not as it should be," and "write so a human can read it": short, direct sentences (one idea each), plain words, active voice, real verbs over nominalizations, and short bulleted lists or table rows instead of piled-up noun stacks. Target the reading level of good developer documentation, not a research paper — accuracy first, but never at the cost of plain prose.

The deliverable deliberately mixes both kinds of content in one document:

  • Unstructured — plain prose and short bulleted lists carrying the reasoning (scope, adversary model, properties provided / not provided, false friends, downstream responsibilities, known misuses).
  • Structured — meaningful tables embedded inline: the §1.7 per-input-operand input-trust table and contract-dimension matrix, the §1.8 output-taint statements, the §1.17 closed disposition set, and the §1.1 boxed triager quick-start. (The separate machine-readable threat-model.yaml is threat-model-sidecar's job.)

Assemble from the upstream artifacts

  • §1.2/§1.3 from the recon component-family carve and out-of-scope inventory.
  • §1.4/§1.5/§1.7/§1.8 from the surface analysis (reachability preconditions, side-effects inventory, per-input-operand table, contract-dimension matrix, output taint).
  • §1.6/§1.9/§1.10/§1.11/§1.12/§1.13/§1.14/§1.15 seeded from recon's mined maintainer positions and promoted as interview answers arrive.
  • §1.1 header, §1.16, §1.17, §1.18 authored here to bind the whole together.
  • Prior-policy back-map appendix from recon whenever SECURITY.md or an equivalent authoritative model existed; retain every source claim until a maintainer explicitly approves removing the map.

Provenance discipline (non-negotiable)

  • Every non-trivial claim carries exactly one of (documented, source), (maintainer, YYYY-MM), (assumption, QN), (inferred, QN). The QN on an assumption or inferred tag resolves to §1.18. No hedge-tags ("implicit", "documented in purpose", "generally known").
  • Prefer documented disclaimers to open questions. When code + docs show a guarantee is simply not made (no thread-safety, no resource bound, no failure atomicity), record it as a documented §1.12 disclaimer — the absence is verifiable. Reserve unresolved / inferred for dimensions where a guarantee plausibly exists but was not confirmed. This is the main lever for cutting MODEL-GAP without weakening closure safety. It is also the easiest thing in the model to overdo, so keep it inside the three bounds in output-structure.md: an absent guarantee is a disclaimer, but an absent behaviour you scanned for is a §1.5 assumption; a disclaimer carries only as far as the source you cite actually reaches; and a guarantee "plausibly exists" when the project has historically fixed reports of that class. Give every disclaimer a boundary and a tier — triage fails closed on a missing tier, so an untiered disclaimer escalates every report it should have answered.
  • An unratified property — inferred or assumption — never carries a security-critical tier. If the guarantee is unratified, leave the matrix row unresolved and put the choice to the maintainer in §1.18. Publishing an unratified security-critical guarantee is worse than publishing a gap: integrators build on §1.11.
  • Assumption vs inferred. Use (assumption, QN) for a conservative default you are willing to act on now (it may close low-blast-radius reports under the relaxed policy); use (inferred, QN) when the question is genuinely open (escalate-only under every policy). Do not relabel a guess as documented to make it close — that launders the author's inference into the project's authority and is forbidden.
  • Every inferred and assumption tag has a matching §1.18 item that states a proposed answer. Mapping is one-directional: inferred/assumption → question required; extra edge-case/meta questions are allowed.
  • Keep the header's draft-confidence count (documented / maintainer / inferred, plus assumption when used) current, and declare the triage policy (strict default, or relaxed). A draft with no inferred / assumption is fully reviewed or overclaiming; mostly unratified is not ready to publish.
  • Retain tags in the published version — a closed report cites (maintainer, 2025-03), and bare prose is not defensible. Footnotes are fine; keep the chain of authority intact.
  • A citation you cannot fill is fixed by opening the file, never by deleting the column. When a Provenance cell resists — you know the claim is right but cannot name a locator — read the source until you can, or tag the row inferred with a §1.18 question. Dropping the column keeps every claim's authority while removing its evidence, and it is invisible in review: the section still looks complete. §1.7, §1.8, §1.10 and §1.15 all require the column for this reaso
阅读完整来源 ↗