開始使用
AI Software Development Suite

AI Software Development Suite

Expert in end-to-end AI-assisted software delivery, from design engineering and design systems to shipping: design engineering, design system automation, technical documentation, spec-driven planning, polyglot code quality, systematic QA, release auditing, LLM context engineering, and Astro frontend optimization. Built for agent-agnostic workflows across the full 9-skill delivery lifecycle.
#工程#產品#生產力
評分
需要更多評價
已售出
2
使用方式
在 Capafy 上執行
也可用於外部應用程式
由創作者提供
fusion

What's in this suite

This bundle covers the complete AI-assisted software development lifecycle — from design system creation through project documentation, spec-driven planning, code quality enforcement, testing, release gating, context optimisation, and frontend delivery.

Free community version of the Penpot MCP skill available at github.com/ar27111994/penpot-mcp — this listing provides the complete packaged, ready-to-install 9-skill suite with continued updates.

Hosted trial configuration

This hosted version includes a shared freemium LLM router for quick evaluation and light experimentation. For production use, team deployments, sensitive workloads, or predictable capacity, users should bring their own provider keys.

Context7 is included as an optional hosted documentation source for API and library reference lookups used by selected workflows. It enhances reference retrieval but is not required for core operation.

In the hosted version, supported services are preconfigured by the platform and injected into the agent runtime. Users do not need to run local setup commands such as npx ctx7@latest setup unless they are reproducing the workflow in their own local Hermes environment.

The 9 skills

Skill Layer What it does When to use it
Design Systems Engineering Design Engineering Create and audit design systems, UI/UX, motion, accessibility, dashboards, and design-to-code handoffs When product design quality, technical handoff, or accessibility rigor matters
Penpot MCP Design Create, audit, and maintain design systems via AI agent Before any code is written; design token and component work
Documentation Generator Document Generate and maintain all project docs: PRDs, architecture, API refs, READMEs, changelogs Initial documentation and ongoing technical writing
GSD Plan & Build Spec-driven multi-session planning that prevents context rot across sessions Starting a project or fixing degrading agent output
Code Quality Standards Quality Quality hierarchy + pitfalls + security checklist for 19 languages Every write, review, and refactor
CLI Tool QA Test SFDPOT-driven QA with structured P1/P2/P3 bug report and release verdict Before releasing a CLI tool or pipeline
Release Readiness Audit Release Gate Structured READY/NOT READY/BLOCKED verdict on a milestone Before tagging any release
LLM Context Modular Docs Context Modular docs that cut agent token usage 70-96% When agents load too much context
Astro Optimization Frontend Performance, accessibility, Core Web Vitals for Astro sites Optimising any Astro-based frontend

How they work together

Design Engineering and Penpot shape the product surface. Documentation Generator captures specs and operating knowledge. GSD drives multi-session delivery. Code Quality Standards keeps implementation disciplined. CLI Tool QA and Release Readiness Audit enforce quality before shipping. LLM Context Modular Docs keeps context lean across every stage. Astro Optimization handles the web delivery layer when the output is an Astro site.

Required setup: Penpot MCP + GSD slash commands + optional Context7

The suite has three setup tiers:

  1. Penpot MCP (required only for the Penpot MCP skill)
  2. GSD slash commands (optional, but recommended if you want /gsd-* workflow shortcuts instead of natural-language invocation)
  3. Context7 MCP (optional local add-on when you want live library/API doc lookup in your own Hermes environment; the hosted Capafy runtime already preconfigures it)

In addition to setup, several skills in this suite now assume a documentation-first workflow when MCP sources are available:

  • Context7 should be used proactively for third-party library/API documentation, setup, configuration, and code-generation grounding.
  • Astro Docs MCP should be used proactively for Astro-specific APIs, imports, setup, configuration, and code examples.
  • Official live web sources via search/fetch should still be used when the task depends on freshness-sensitive facts like latest versions, release dates, migration notes, deprecations, or recently changed docs.

Penpot MCP key

The Penpot MCP skill (one of the 9 in this bundle) requires a Penpot MCP key to connect to your Penpot instance:

  1. Open Penpot → Main menu → Your account → Access tokens
  2. Click New token, name it (e.g. mcp-agent), copy the token
  3. Your MCP client URL: http://<your-machine>:4401/mcp?userToken=YOUR_TOKEN

GSD slash-command install

The GSD skill works through natural-language prompting out of the box, but the /gsd-* workflow shortcuts require a separate install:

npx gsd-core@latest

That installs the workflow-command layer used by supported environments such as Claude Code and similar agent IDEs. The bundled references/client-install.md guide covers environment-specific paths and manual fallback setup.

Context7 MCP (optional for local Hermes setups)

Context7 is included in the hosted Capafy runtime as an optional documentation source for API and library reference lookups used by selected workflows. If you are reproducing the suite locally in Hermes and want the same capability, configure the Context7 MCP server in your Hermes MCP setup using endpoint https://mcp.context7.com/mcp with a CONTEXT7_API_KEY header. You do not need to run local Context7 setup commands when using the hosted version.

MCP assumptions and freshness policy

  • Context7: use proactively for third-party library/API documentation, setup, configuration, and code-generation grounding when it is available.
  • Astro Docs MCP: use proactively for Astro-specific APIs, imports, setup, configuration, and code examples when it is available.
  • Web search/fetch: still use official live web sources via search/fetch when the task depends on freshness-sensitive public facts such as latest versions, release dates, migration notes, deprecations, or recently changed documentation.
  • Fallback order: MCP docs first for structured reference; official live web sources via search/fetch second for freshness; local repo docs/code/tests for project-specific behavior.
  • If unavailable: if MCP docs and live web tools are both unavailable, say so instead of implying current-version certainty you cannot verify.

Each skill works standalone

All nine skills can be installed and used independently. This bundle packages them together for teams who want the complete workflow.


外部 API

penpot.app/mcp/stream?userToken=YOUR_MCP_KEY · mcp.context7.com/mcp · mcp.docs.astro.build/mcp