agentleFS
crankyjediiicrankyjediii/vibecraft/skills/system-architect/SKILL.md

system-architect

Skill1 starsChanged 7 days agoMITView on GitHub

Phase 3 System Architect for VibeCraft. Designs Mermaid flow and sequence diagrams, strict domain contracts, file tree, and Phase 1 gotcha mitigations inside a draft launchpad_artifact.html before code generation. Activate on Phase 3 or /vibecraft:launchpad.

32 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use database-architect, webapp-testing and 4 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • xfstudioSkill

    database-architect

    Expert database architect specializing in data layer design from

    Also in 3 projects that use this

  • JetBrainsSkill

    webapp-testing

    Toolkit for interacting with and testing local web applications using Playwright. Supports verifying frontend functionality, debugging UI behavior, capturing browser screenshots, and viewing browser logs.

    Also in 4 projects that use this

  • Microsoft AzureSkill

    git-workflow

    Squad branching model: dev-first workflow with insiders preview channel

    Also in 3 projects that use this

  • BlockSkill

    frontend-design

    Create distinctive, production-grade frontend interfaces with high design quality. Use this skill when the user asks to build web components, pages, or applications. Generates creative, polished code that avoids generic AI aesthetics.

    Also in 5 projects that use this

  • MicrosoftSkill

    mcp-builder

    Guide for creating high-quality MCP (Model Context Protocol) servers that enable LLMs to interact with external services through well-designed tools. Use when building MCP servers to integrate external APIs or services, whether in Python (FastMCP) or Node/TypeScript (MCP SDK).

    Also in 3 projects that use this

  • SentrySkill

    skill-creator

    Create new agent skills following the Agent Skills specification. Use when asked to "create a skill", "add a new skill", "write a skill", "make a skill", "build a skill", or scaffold a new skill with SKILL.md. Guides through requirements, writing, registration, and verification.

    Also in 5 projects that use this

The file

We split this skill into the rules it gives an agent

5 rules. 0 are common, 0 are distinctive, and 5 are very specific to this project.

  • 0

    Common

    Most of the 28 edited versions we read keep these.

  • 0

    Distinctive

    This author's own choices. Worth a look.

  • 5

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Every Phase 1 [Gotcha] mitigation maps to at least one named module and contract type in the blueprint.

    line 21
  2. Very specific:

    Design minimal module boundaries (LAUNCH.tree): Keep MVP module count between 5 and 14 files.

    line 25
  3. Very specific:

    Mermaid readability cap: Keep flow to <= 12 nodes and sequence to <= 6 participants and <= 14 messages so diagrams render legibly at 768px …

    line 39
  4. Very specific:

    Do not start Phase 4 code generation until the user replies 1, "go", "approve", or approves a specific revision with "do X and build".

    line 77
  5. Very specific:

    If a Phase 1 [Gotcha] contradicts a new user chat override (for example user asks to remove the Worker while keeping synchronous OPFS …

    line 78

“Common” is measured: kept by at least 6 in 10 of the 28 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Phase 3: System Architect

Purpose

Convert the locked stack (VIBECRAFT PICK) and design tokens (VIBECRAFT LOCK) into an explicit, reviewable architectural blueprint (vibecraft/03-architecture.html) containing the module file tree, Mermaid component flow and sequence diagrams, strict domain type contracts, and concrete mitigations for both Phase 1 gotchas before a single line of application logic is written.

Inputs

  • Locked Phase 1 stack + two [Gotcha] items (VIBECRAFT PICK).
  • Locked Phase 2 :root CSS custom properties (VIBECRAFT LOCK), unless skipped for a headless CLI/API build.
  • Code quality rules in ../vibecraft/references/code-quality.md.

Preconditions

  1. Phase 1 and Phase 2 gates passed (or standalone /vibecraft:launchpad invoked on an existing repository).
  2. Every Phase 1 [Gotcha] mitigation maps to at least one named module and contract type in the blueprint.

Step Procedure

  1. Design minimal module boundaries (LAUNCH.tree): Keep MVP module count between 5 and 14 files. Separate pure domain logic and data adapters from UI components so business rules remain unit-testable without a DOM.
  2. Author strict domain contracts (LAUNCH.contracts): Define every shared entity, state union, storage repository interface, and typed Result<T, AppError> boundary. Zero any, zero untyped Record<string, unknown> where fields are known. Set LAUNCH.contractsLang ('ts', 'rust', 'python', 'go', or 'sql') to match the locked stack.
  3. Draw Mermaid system & sequence diagrams (LAUNCH.flow, LAUNCH.sequence):
    • flow: flowchart TD or flowchart LR (<= 12 quoted nodes) showing UI, state/worker boundary, storage, and external I/O. Quote every node label containing parentheses, slashes, or colons (A["Worker (OPFS SQLite)"]).
    • sequence: sequenceDiagram tracing the single riskiest end-to-end write or failure-recovery path, explicitly showing the Phase 1 [Gotcha] mitigation in action.
  4. Emit draft launchpad_artifact.html: Read templates/generative-ui/launchpad_artifact.html and populate LAUNCH with phase: 'Phase 3 of 4 · Architecture Draft', title: 'Review the architecture', name, tagline, contractsLang, tree, flow, sequence, and contracts (setting readme, commands, logo, midjourney, dalle, nextSteps to concise draft placeholders noted as finalized in Phase 4). Write to vibecraft/03-architecture.html in Claude Code or render in Claude Desktop.
  5. Reply in chat + approval gate: Emit the 3-line HUD (Phase 3/4 · Architecture review · Reply 1 to build, or 2/3 to revise), 1 to 2 prose lines naming the single highest-risk architectural trade-off, and the gate micro-prompt:
    Gate · Blueprint: [1] Approve and build*  [2] Change the flow  [3] Change the data model

Decision Rules (Numeric Thresholds)

  • File-tree complexity cap: Weekend MVPs (Q2=1) use 5 to 10 source files; production v1 (Q2=2) uses 8 to 16 source files. Never propose empty utils/ grab-bags or > 3 directory nesting levels without justification.
  • Mermaid readability cap: Keep flow to <= 12 nodes and sequence to <= 6 participants and <= 14 messages so diagrams render legibly at 768px and 1440px without horizontal clipping.
  • Gotcha traceability (100%): 100% of [Gotcha] items from VIBECRAFT PICK must appear as explicit types, worker boundaries, or middleware steps in contracts and sequence.

Output Contract

interface LaunchpadData {
  phase: string;          // 'Phase 3 of 4 · Architecture Draft' (or 'Phase 4 of 4 · Launchpad' in Phase 4)
  title: string;          // 'Review the architecture' (or '<Name> is ready to launch' in Phase 4)
  name: string;           // Non-empty product name, e.g. 'LedgerSlip'
  tagline: string;        // 1-sentence value proposition
  contractsLang: 'ts' | 'rust' | 'python' | 'go' | 'sql';
  readme: string;         // Draft summary in Phase 3; complete runbook in Phase 4
  commands: string;       // Exact install/dev/test commands
  tree: string;           // ASCII file tree with 1-phrase purpose per file
  flow: string;           // Valid Mermaid flowchart TD / LR (quoted labels)
  sequence: string;       // Valid Mermaid sequenceDiagram
  contracts: string;      // Compilable strict type definitions
  logo: string;           // Geometric SVG string (2 to 5 primitives, viewBox="0 0 64 64")
  midjourney: string;     // Optional brand imagery prompt (finalized in Phase 4)
  dalle: string;          // Optional icon exploration prompt (finalized in Phase 4)
  nextSteps: string[];    // 3 to 5 concrete post-MVP milestones
}

Worked Examples

See references/examples.md for two complete runs:

  1. Clean Approval Flow: Local-first OPFS SQLite note app showing SharedWorker lock serialization in sequence and strict Result<Note, StorageError> contracts.
  2. Revision Loop (User replies "3" without details): Elicits the exact schema fork in 1 targeted micro-prompt, updates LAUNCH.contracts and LAUNCH.sequence, and re-presents [1]* [2] [3].

Failure Modes & Recovery

  • User replies 2 (Change the flow) or 3 (Change the data model) with no explanation: Do not guess blindly and do not advance to Phase 4. Ask 1 targeted 1-line question or 3-option micro-prompt naming the main design fork, update vibecraft/03-architecture.html, and re-emit Gate · Blueprint: [1] Approve and build* [2] Change the flow [3] Change the data model.
  • Mermaid syntax hazard: Unquoted parentheses Node[Worker (OPFS)] crash Mermaid's parser. Always wrap labels containing (, ), [, ], :, or / in double quotes: Node["Worker (OPFS)"].

Refusal & Escalation Rules

  • Do not start Phase 4 code generation until the user replies 1, "go", "approve", or approves a specific revision with "do X and build".
  • If a Phase 1 [Gotcha] contradicts a new user chat override (for example user asks to remove the Worker while keeping synchronous OPFS SQLite), state the browser spec blocker (FileSystemSyncAccessHandle throws on main thread [Verified]) in 1 line and offer [1] Keep dedicated Worker* [2] Switch to async IndexedDB.

Token / Word Budget

  • SKILL.md target: < 1,100 words (< 5,800 bytes).
  • Chat reply budget: 3-line HUD + 1 to 2 prose lines (<= 45 words) + 1-line Gate · Blueprint micro-prompt.

Self-Check List

  • Are LAUNCH.name, LAUNCH.tagline, LAUNCH.phase, LAUNCH.title, and LAUNCH.contractsLang all non-empty?
  • Are all Mermaid node labels containing special characters wrapped in "..."?
  • Does LAUNCH.contracts have zero any types and explicit error union types for every I/O method?
  • Are both Phase 1 [Gotcha] mitigations visible in tree, sequence, and contracts?

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.