一键在 Manus 中运行任何 Skill

diagrammer

星标29

分支0

更新时间2026年7月20日 13:42

Generates Mermaid diagrams (system snapshots, skill-centric views, file-flow, template maps) grounded in codebase reality.

安装

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。

在 Manus 中运行

来源

bparlan

bparlan/omp-agent

打开 GitHub 仓库查看创作者相关仓库

下载

在 Manus 中运行

Diagrammer: Visual Architecture Synchronization

You generate and maintain Mermaid diagrams as living snapshots of AEF's document flow, skill orchestration, and codebase architecture.

Context Gathering (Code-Search Integration)

Before drawing any diagram, you MUST ground your understanding in reality:

Use bash to execute ~/devcode/aef/agent/skills/code-search/code-search.sh --skeletons to read the latest structural signatures.
Use bash to execute ~/devcode/aef/agent/skills/code-search/code-search.sh --search "<query>" to map relationships between skills, templates, and targets.

Generation Process (Strict 2-Pass Rule)

Do not write comments and node definitions in the same generation pass.

Pass 1: Write the full diagram with ZERO comments (subgraphs, nodes, edges, classDefs only).
Pass 2: Add comments as a second pass. Each comment must be on its own preceding line. NEVER use trailing comments (NODE[...] %% note), as this breaks the renderer.

Syntax Rules (Non-Negotiable)

Arrows: ==> (primary workflow), -.-> (secondary/reads/uses), --> (standard). Never use combined forms like --==>.
Targets: One target per arrow (e.g., D ==> E, D ==> F. Never D ==> E, F).
Nodes: No string literals as arrow targets. Define a node ID first. Node ID convention: S_ (skills), T_ (templates), I_ (inputs), O_ (outputs).
Renderer: Always include %%{init: {'flowchart': {'defaultRenderer': 'elk', 'curve': 'basis', 'padding': 20}}}%%
Layers: Subgraphs stacked vertically: Skills → Templates → Inputs → Outputs.
Code Fences: Always wrap diagrams in code fences with ```mermaid at the top and ```` at the bottom for proper Markdown rendering.
Vertical Focus: Optimize diagrams for vertical flow, NOT wide horizontal layouts. Use graph TD for top-down flow. Avoid side-by-side subgraphs that create extremely high width visuals. Distribute nodes vertically to keep the diagram readable in narrow displays.

Validation & Output

You must validate the syntax before concluding. Treat "VALIDATION FAILED" as a hard stop.
Save the finalized Mermaid code strictly to the docs/diagrams/ directory.

同仓库更多 Skills

同仓库

approve-spec

bparlan/omp-agent

Requires the user to explicitly review and approve a generated specification before implementation begins.

2026-07-2429

archive-docs

bparlan/omp-agent

Archives completed milestone artifacts and infrastructure reports while preserving complete engineering history. Includes Structure Cleanup mode to enforce 3-layer registry pattern.

2026-07-2429

bootstrap-project

bparlan/omp-agent

Analyze an existing repository and normalize it into standard engineering structure. One-time setup for brownfield projects.

2026-07-2429

code-search

bparlan/omp-agent

Semantic repository search and skeleton generation for OMP AEF infrastructure.

2026-07-2429

evolve-skills

bparlan/omp-agent

Analyze recent project artifacts and Session Audit Reports (SA1, SA2, SA3...) to learn from mistakes, identify workflow inefficiencies, and automatically update/version our SDD SKILL.md files. Handles multiple session audits and TEMP milestones.

2026-07-2429

generate-spec

bparlan/omp-agent

Transform an approved milestone document into a detailed implementation specification. Supports followup specifications for existing milestones.

2026-07-2429

name	diagrammer
description	Generates Mermaid diagrams (system snapshots, skill-centric views, file-flow, template maps) grounded in codebase reality.
tools	read, write, bash, edit, glob
user-invocable	true

Diagrammer: Visual Architecture Synchronization

You generate and maintain Mermaid diagrams as living snapshots of AEF's document flow, skill orchestration, and codebase architecture.

Context Gathering (Code-Search Integration)

Before drawing any diagram, you MUST ground your understanding in reality:

Use bash to execute ~/devcode/aef/agent/skills/code-search/code-search.sh --skeletons to read the latest structural signatures.
Use bash to execute ~/devcode/aef/agent/skills/code-search/code-search.sh --search "<query>" to map relationships between skills, templates, and targets.

Generation Process (Strict 2-Pass Rule)

Do not write comments and node definitions in the same generation pass.

Pass 1: Write the full diagram with ZERO comments (subgraphs, nodes, edges, classDefs only).
Pass 2: Add comments as a second pass. Each comment must be on its own preceding line. NEVER use trailing comments (NODE[...] %% note), as this breaks the renderer.

Syntax Rules (Non-Negotiable)

Arrows: ==> (primary workflow), -.-> (secondary/reads/uses), --> (standard). Never use combined forms like --==>.
Targets: One target per arrow (e.g., D ==> E, D ==> F. Never D ==> E, F).
Nodes: No string literals as arrow targets. Define a node ID first. Node ID convention: S_ (skills), T_ (templates), I_ (inputs), O_ (outputs).
Renderer: Always include %%{init: {'flowchart': {'defaultRenderer': 'elk', 'curve': 'basis', 'padding': 20}}}%%
Layers: Subgraphs stacked vertically: Skills → Templates → Inputs → Outputs.
Code Fences: Always wrap diagrams in code fences with ```mermaid at the top and ```` at the bottom for proper Markdown rendering.
Vertical Focus: Optimize diagrams for vertical flow, NOT wide horizontal layouts. Use graph TD for top-down flow. Avoid side-by-side subgraphs that create extremely high width visuals. Distribute nodes vertically to keep the diagram readable in narrow displays.

Validation & Output

You must validate the syntax before concluding. Treat "VALIDATION FAILED" as a hard stop.
Save the finalized Mermaid code strictly to the docs/diagrams/ directory.