Skip to content

A Claude Code plugin that provides a conversational, adaptive, and agentic subagent for co-creating rigorous, multi-perspective technical specs—featuring dynamic memory graphs, adaptive workflows, and explicit multi-role analysis. Especially useful for non-technical users and teams adopting the Spec Kit Methodology.

License

Notifications You must be signed in to change notification settings

jcmrs/claude-code-spec-kit-subagent-plugin

Repository files navigation

Claude Code Spec Kit Subagent Plugin

Spec Kit Partner is now provided as a Claude Code plugin: a next-generation, conversational, adaptive, and agentic tool designed to help humans co-create rigorous, multi-perspective technical specs for GitHub Spec Kit projects. It leverages dynamic memory graphs, adaptive workflows, and explicit multi-role analysis to bridge human creativity and technical rigor. Yes, it is particularly useful for humans who are non-technical and/or not trained yet in Spec Kit Methodology.


Features

  • Conversational Orchestration: Natural, empathic dialogue management tailored to user preferences.
  • Multi-Role Analysis: Dynamically identifies and embodies relevant professional perspectives for robust analysis.
  • Memory Graph: Persists all insights, relationships, and synthesis steps in a queryable, file-based graph.
  • Adaptive Workflow: Adjusts process phases and requirements in response to user input and context.
  • Transparent Logging: All actions, analysis, and conversations are logged for auditability and recovery.
  • Claude Code Marketplace Compatible: Fully supports Claude Code plugin marketplaces, including a complete .claude-plugin/marketplace.json descriptor.

Plugin Directory & File Structure

All plugin metadata, code, and agent-generated artifacts are organized as follows:

claude-code-spec-kit-subagent-plugin/
├── .claude-plugin/
│   ├── plugin.json
│   └── marketplace.json
├── agents/
│   └── spec-kit-partner.md
├── spec-kit-partner/
│   ├── src/
│   │   ├── main.py
│   │   ├── workflow_manager.py
│   │   ├── memory_graph.py
│   │   ├── user_profile.py
│   │   └── multi_role_analysis.py
│   └── project-data/
│       ├── memory-graph.json
│       ├── user_profile.json
│       ├── workflow_state.json
│       ├── multi-role-analysis.json
│       ├── logs/
│       │   ├── conversation.log
│       │   └── internal_monologue.log
│       ├── spec/
│       │   ├── spec.md
│       │   └── state.json
│       └── diagrams/
│           ├── system_flow.mermaid
│           ├── memory_graph.mermaid
│           ├── workflow_state_machine.mermaid
│           ├── multi_role_analysis_flow.mermaid
│           ├── agent_user_interaction.mermaid
│           ├── data_flow.mermaid
│           └── role_perspective_map.mermaid
├── assets/
│   ├── icon.png
│   ├── screenshot1.png
│   └── screenshot2.png
├── README.md
├── LICENSE
└── CHANGELOG.md

Marketplace Metadata

Marketplace and plugin metadata are defined in:

  • .claude-plugin/plugin.json — Main plugin manifest.
  • .claude-plugin/marketplace.json — Marketplace listing fields including display_name, summary, categories, icon, documentation_url, publisher, license, screenshots, and more.

See Official Claude Code Marketplace Docs.


Diagram Output Types

All diagrams are output as Mermaid (.mermaid) format for easy rendering and AI parsing. Rendered .svg or .png versions may also be exported, but .mermaid is canonical.


Initialization Protocol

On first run or when required files are missing:

  • Ensure all required directories exist before writing.
  • Generate all support code and persistent data files as described in Initialization Protocol.
  • Populate each .json file using provided schemas and set initial values.
  • Log all steps and errors to project-data/logs/conversation.log.

See spec-kit-partner.md for full protocol.


Usage

  1. Install this plugin into your Claude Code environment using your plugin manager or by cloning this repository under /plugins/.
  2. On first invocation, the plugin will create and initialize all required files and directories in its workspace.
  3. All runtime state, logs, diagrams, and outputs are self-contained within the plugin directory.
  4. For more details on engine architecture, data schemas, and interaction protocols, refer to the main spec file or plugin documentation.

Development & Contributing

This repository includes automated workflows to ensure code quality:

  • Python Linting - Validates syntax and style of Python code
  • JSON Validation - Ensures plugin manifests are valid and complete
  • Asset Validation - Checks that referenced files exist
  • Markdown/Mermaid Linting - Validates documentation syntax
  • Python Tests - Runs automated test suite

All workflows run automatically on pull requests. See CONTRIBUTING.md for detailed development guidelines.


License

This project is licensed under the MIT License.


Contact & Support

  • For questions, contributions, or collaboration, open an issue or contact the repository maintainer.
  • For support, visit: GitHub Issues

For questions, contributions, or collaboration, open an issue or contact the repository maintainer.

About

A Claude Code plugin that provides a conversational, adaptive, and agentic subagent for co-creating rigorous, multi-perspective technical specs—featuring dynamic memory graphs, adaptive workflows, and explicit multi-role analysis. Especially useful for non-technical users and teams adopting the Spec Kit Methodology.

Topics

Resources

License

Contributing

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Contributors 2

  •  
  •