ContextShare (VS Code Extension)

September 23, 2025 · View on GitHub

Unified AI workflow catalog for VS Code: manage and share chat modes, instructions, prompts, tasks - plus upcoming MCP server orchestration - across all your repositories from one consistent UI. Reduce prompt drift, standardize team AI usage, and reuse curated presets.

✨ Feature Highlights

Core value: a single, structured, multi-catalog layer for AI assistant resources that become instantly available & activatable per workspace.

Catalog aggregation (local folders + remote URLs)
Chat Modes (persona definitions)
Instructions (reusable guidance bundles)
Prompts (templated starting points)
Tasks (JSON task configs)
MCP Integration (coming soon)
Presets (presets bundling multiple resources)
Real‑time sync & state tracking (INACTIVE / ACTIVE / MODIFIED)
Safe activation (never overwrites user‑created originals)
Secure: HTTPS-only remotes + path & filename sanitization

Feature Matrix

Category	Purpose / Scope	Runtime Location	Filename Pattern / Notes	Status
Chat Modes	Persona / behavior definitions	`.github/chatmodes/`	`*.chatmode.md`	✅
Instructions	Shared guideline sets	`.github/instructions/`	`.instructions.md` (legacy .instruction.md)	✅
Prompts	Prompt templates / starters	`.github/prompts/`	`*.prompt.md`	✅
Tasks	Automation / action configurations	`.github/tasks/`	`*.task.json`	✅
MCP Servers	Model Context Protocol sources	`.vscode/mcp.json`	Merged composite file	Coming soon
Presets	Declarative preset bundles	`.github/presets/`	`*.json` (see example below)	✅

State logic: ACTIVE resources are copied to runtime; MODIFIED indicates the runtime file diverged from its source (e.g., team-local customization).

🚀 Quick Start

Installation

Install from the VS Code Marketplace: ContextShare

Use in 5 Steps

Open the ContextShare Activity Bar view.
Add a catalog (Options → Add Catalog Directory… or add a remote HTTPS URL).
Activate a resource (right‑click → Activate).
(Optional) Apply a Preset to activate multiple at once.
Edit runtime copies under .github/** if you need local tweaks (they'll show as MODIFIED).

Presets

Presets are small JSON descriptors bundling chosen chat mode + instructions + prompts + tasks (+ soon MCP servers) into a one‑click activation set. Great for role or workflow switching (e.g., "Full Stack Review", "Security Audit").

Example Catalog Structure

example-catalog/
├── chatmodes/
│   └── developer-assistant.chatmode.md
├── instructions/
│   └── code-review-guidelines.instructions.md
├── prompts/
│   └── bug-analysis.prompt.md
├── tasks/
│   └── automated-testing.task.json
├── mcp/
│   └── development-servers.mcp.json
└── presets/
    └── full-stack-dev.json

📖 Documentation

Setup Guide - Detailed installation and configuration instructions
Software Architecture - Technical architecture overview
Test Plan - Testing strategies and test cases
Changelog - Version history and release notes
Catalog Display Names - Naming conventions and examples

Configuration Examples

Add catalogs in .vscode/settings.json:

{
    "contextshare.catalogs": [
        "../team-ai-catalog", 
        "https://raw.githubusercontent.com/example-org/shared-ai-catalog/main/index.json"
    ]
}

Remote index.json (HTTPS only):

{
    "version": 1,
    "resources": [
        "prompts/bug-analysis.prompt.md",
        "instructions/code-review-guidelines.instructions.md",
        "chatmodes/developer-assistant.chatmode.md"
    ]
}

Sample Preset (.github/presets/full-stack-dev.json):

{
    "name": "Full Stack Dev",
    "chatMode": "developer-assistant.chatmode.md",
    "instructions": ["code-review-guidelines.instructions.md"],
    "prompts": ["bug-analysis.prompt.md"],
    "tasks": ["automated-testing.task.json"],
    "mcpServers": ["development-servers.mcp.json"]
}

Display Names: Provide friendly catalog labels via configuration so multiple catalogs can be visually distinguished in the tree (see CATALOG_DISPLAY_NAMES_EXAMPLE.md).

🛠️ Development

Development details, build, and contribution workflow are documented in CONTRIBUTING.md and supporting architecture docs.

🤝 Contributing

We welcome contributions! Please see our Contributing Guidelines for details.

Development Process

Fork the repository
Create a feature branch: git checkout -b feature/amazing-feature
Write tests for your changes
Implement your feature
Ensure all tests pass: npm test
Commit your changes: git commit -m 'Add amazing feature'
Push to the branch: git push origin feature/amazing-feature
Open a Pull Request

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.

🔒 Security & Privacy

Focused on safe local workspace operations:

HTTPS-only remote catalog sources
Path traversal & filename sanitization
User-created runtime resources are never deleted
No telemetry / tracking collected by this extension

Resources:

Report vulnerabilities privately via the Security Policy (responsible disclosure). Please avoid public issues for sensitive findings.

❓ Troubleshooting

Issue	Possible Cause	Action
Resource not listed	Filename pattern mismatch	Verify suffix (e.g., `.prompt.md`, `.instructions.md`)
Remote catalog empty	`index.json` not reachable / HTTP error	Open URL in browser; ensure HTTPS and correct raw path
Resource shows MODIFIED unexpectedly	Local edit vs source catalog	Diff runtime file in `.github/**` with original source
Removed MCP server persists	(Upcoming MCP feature) cached merged entry	After MCP feature release: remove from source & reload window
Preset not applying all items	Missing referenced filenames	Check JSON fields and ensure each resource exists

If stuck, enable verbose logging (future setting) or open an issue with a minimal reproduction.

🏷️ Versioning

We use Semantic Versioning for release management:

MAJOR: Breaking changes
MINOR: New features (backward compatible)
PATCH: Bug fixes (backward compatible)

Code of Conduct

This project adheres to the Contributor Covenant Code of Conduct. Participation implies acceptance.

Trademarks

All product names, logos, and brands are property of their respective owners. Use of any third-party trademarks or logos does not imply endorsement.