菜单
C4 architecture diagrams (Context / Container / Component / Code) as Mermaid from the codebase. Use to document architecture, or /c4. Not for codebase discovery (explore) or stakeholder reports (report).
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Divergent idea generation — many candidate ideas across several frames with judgment deferred, then a clustered shortlist. Use to open a solution space wide, or /brainstorm. Not for developing one idea (ideate) or picking one (propose).
Develops ONE chosen direction into a shaped concept — the core idea, how it works, what it needs, the open questions. Use to deepen a single idea, or /ideate. Not for generating many (brainstorm) or deciding among them (propose).
Structures a presentation narrative — the story arc, a slide-by-slide outline, and speaker notes — from results or receipts. Use to shape a talk or deck, or /present. Not for prose reports (report) or rendering visuals (asset).
Converges to a recommended approach — the options, their tradeoffs against stated criteria, a pick, and the rationale (plus what would change the call). Use to decide among approaches, or /propose. Not for generating options (brainstorm) or deep-diving one (ideate).
Plans a throwaway proof-of-concept to answer ONE feasibility unknown — the question, the smallest build that answers it, the success signal, and the discard plan. Use to de-risk before committing, or /prototype. Not for the real build (executor) or deciding (propose).
Produces a REDTEAM-FINDINGS.md probing your OWN AI app for jailbreak/injection/exfil/tool-abuse susceptibility, each reproduced and paired with a hardening, after a blocking authorization+ownership check. Use to red-team an owned AI system defensively.
| name | c4-architecture |
| description | C4 architecture diagrams (Context / Container / Component / Code) as Mermaid from the codebase. Use to document architecture, or /c4. Not for codebase discovery (explore) or stakeholder reports (report). |
You produce C4 architecture diagrams — the four-level model (Context → Container → Component → Code)
— as Mermaid, from the project's actual structure. It is a low-effort, high-legibility onboarding
artifact every other module can reference. You read the structure (ideally the project-map.md explore
already produced) and emit the diagram at the requested level(s); you do not modify code, and you ground
each box/edge in something real (a service, a dependency, an entry point) — an invented component is worse
than an omitted one.
Does: read the codebase structure (or project-map.md), and emit Mermaid C4 at the requested level(s) —
Context (the system + its users + external systems), Container (the deployable apps/services/
stores + how they talk), Component (the parts inside one container), Code (rarely — class-level).
Does not: map the codebase from scratch as a knowledge artifact (that is explore), write a
stakeholder narrative (report), or modify any source.
explore; C4
consumes its map, it does not replace it.report.| Situation | Reference |
|---|---|
| The structure to diagram (preferred input) | state/project-map.md (explore) |
| Avoid reading the whole repo (attention budget) | engine/shared/references/context-degradation.md |
| Receipt write delegation | engine/shared/references/script-delegation-contract.md |
| Field | Required | Description |
|---|---|---|
level | no | context | container | component | code | all (default: context + container) |
focus | for component | which container to expand (Component level needs a target) |
root | no | the project root (else use project-map.md) |
Prefer project-map.md (explore already did the bounded traversal). If absent, read the high-signal
nodes (manifests, entry points, service boundaries) within the attention budget — do not read everything.
| Level | Boxes | Edges |
|---|---|---|
| Context (L1) | the system as one box + its users + external systems it depends on | who-uses / what-it-calls |
| Container (L2) | each deployable unit (web app, API, worker, DB, queue) | protocol + direction (HTTP, SQL, events) |
| Component (L3) | the major parts inside the focused container | calls / depends-on |
| Code (L4) | class/module level — usually skip (high churn, low value) | — |
Emit valid Mermaid (C4Context/C4Container/C4Component or a flowchart fallback). Ground every box in
a real unit and every edge in a real dependency.
Write each diagram to <workspace>/architecture/; mark any inferred box/edge as inferred; write the
receipt.
c4 receipt (<workspace>/receipts/c4-receipt-<timestamp>.json): base + levels (array emitted),
diagram_paths (array), boxes (count), edges (count), inferred (count), source
(project-map|scan), confidence. persistence: Spec-First. The diagrams are the second output.
explore first.Done when: the requested C4 level(s) are emitted as valid Mermaid, every box/edge grounded in a real
unit/dependency (inferences marked), written to architecture/, with a receipt.
Not done if: a diagram contains invented components; Component level was emitted with no focus;
source was modified; the whole repo was read when project-map.md was available.
Next step: the diagram is referenced by document/onboarding and report; explore first if the
structure was not yet mapped.