PipelineKit：可移植的Claude Code工作流框架与多代理开发套件

PipelineKit is an open-source project developed by Bruce188, designed as a portable, sandbox-ready workflow overlay for Claude Code. Key features include:

• 25 native skills covering the full development lifecycle
• 14 specialized agents for tasks like architecture design and code review
• 21 production-grade Git hooks for automated quality assurance
• Full integration with MCP (Model Context Protocol) and LSP (Language Server Protocol)
• One-click deployment via devcontainer or GitHub Codespaces

Its core goal is to enable developers to quickly set up a complete AI-integrated development environment through simple pull, install, and run commands.

PipelineKit was created by Bruce188 to address the need for a reproducible, portable AI-assisted development workflow. The design philosophy emphasizes:

• Deep integration of AI capabilities into the development process
• Maintaining environment reproducibility and portability
• Enabling quick setup of multi-agent pipelines, MCP tool clusters, LSP protocol stacks, and TDD-aware pipelines with minimal configuration.

PipelineKit supports multiple deployment options:

Recommended: GitHub Codespaces Open the repository in GitHub Codespaces; the devcontainer's post-create hook automatically runs the installation script, requiring no local setup.

Local Installation

1. Clone the repository: git clone https://github.com/Bruce188/pipelinekit.git
2. Navigate to the directory: cd pipelinekit
3. Run the idempotent install script: ./scripts/install.sh

After installation, start the configured Claude Code environment with the claude command.

PipelineKit uses a layered architecture:

Rules Layer Defines Claude Code's behavior via files like CLAUDE.md (main config), rules/workflow.md (development流程规范), and rules/agents-worktrees.md (agent worktree management).

Skills Layer 25 native skills covering analysis (analyze, review), planning (create-plan, implement-plan), and workflow management (pipeline, ppr).

Agent Layer 14 specialized agents including architect (system design), code-reviewer (quality checks), security-auditor (vulnerability detection), tdd-test-writer (test case generation), and tdd-implementer (code implementation).

Hooks Layer 21 Git hooks for quality assurance, such as validate-commit-msg (format checks), strip-ai-attribution (remove AI署名), block-push-main (prevent direct main branch pushes), and tdd-order-check (ensure test-first workflow).

MCP Integration PipelineKit integrates with MCP tools like context7 (context-aware code understanding), serena (semantic search), sequential-thinking (complex reasoning), and optional local-rag (private codebase RAG).

LSP Integration Supports LSP servers for multiple languages: pyright (Python), TypeScript/JavaScript, C#, Go (gopls), and Rust (rust-analyzer).

TDD Workflow via /pipeline Command The /pipeline command routes tasks based on docs/features.md prefixes:

• Dev Mode (TDD Enabled) : For feat/fix/refactor/perf/test tasks; auto-generates tests via tdd-test-writer then implements code via tdd-implementer.
• Non-Dev Mode: For docs/chore/style/build/ci etc. tasks; skips TDD. Override via Type: dev|non-dev in docs/features.md.

Memory System PipelineKit builds project-specific memory in ~/.claude/projects/<project-slug>/memory/, following the schema in claude/memory/MEMORY.md to maintain cross-session context.

Personalization: Caveman Wenyan-Ultra Mode This mode uses concise, direct language (omitting articles and filler words). Switch modes with /caveman lite | full | ultra or revert to normal with stop caveman.

Security Design

• No subprocess-driven scripts like orchestrate.sh; all workflows managed via in-process /pipeline skills.
• Avoids claude -p subprocess calls to reduce security risks.
• Templated path handling: strips memory data and clears email addresses.
• No auto-installation of market plugins; manual installation allowed after setup.

Extensibility

• gstack Overlays: Add features via /gstack-* skills.
• Custom Skills: Add local skills in ~/claude-skills/.
• Template Resources: Use tresor-resources for prompt templates and standards.

PipelineKit provides a complete solution for teams integrating AI into their development workflows. It is ideal for:

• Teams needing standardized AI-assisted development processes
• Projects enforcing strict TDD practices
• Environments requiring consistent AI behavior across multiple setups
• Organizations with high code quality and security requirements

By offering a portable, extensible, production-ready framework, PipelineKit lowers the barrier to adopting AI-assisted development with Claude Code.