skills: The Official Upstash Skills Bundle for DeepSeek Harness
upstash/skills
Upstash skills and MCP server for DeepSeek Harness, covering Redis, vector, rate limiting, and more.
Provides standardized skills for the Upstash product suite to DeepSeek Harness, letting AI agents directly use Redis, Vector, QStash, and more.
dsh plugin --profile web add github:upstash/skills- Category
- Development & Operations
- Platform
- DSH-Plugin
- Author
- upstash
- Distribution
- Plugin
skills Key Features
skills Repository
skills Repo Summary
Upstash Agent Skills is an official skills bundle (DSH plugin) for DeepSeek Harness, maintained by the Upstash team under the MIT license. It provides a collection of skills that extend AI coding agents' capabilities when working with Upstash SDKs. This plugin solves the problem of lacking standardized, reusable instructions for integrating Upstash services like Redis, QStash, and Vector, by packaging instructions and resources that agents can leverage. It includes skills for Redis caching, full-text search, vector databases, rate limiting, serverless messaging, sandboxed cloud containers, and more, along with an MCP server for direct API access. As a DSH plugin, it also works with Claude Code, Cursor, OpenAI Codex, and other agent platforms, and allows credential configuration via the /upstash-login command within sessions.
What core skills does the Upstash Agent Skills DSH plugin include?
This DSH plugin bundles 11 skills covering the Upstash product suite: upstash is a combined skill for all SDKs; upstash-redis-js provides serverless Redis for caching, sessions, leaderboards, and full-text search; upstash-vector-js covers vector database features and SDK usage; upstash-qstash-js enables serverless messaging and scheduling via HTTP endpoints; upstash-ratelimit-js offers rate limiting with the Redis Rate Limit TypeScript SDK; upstash-search-js focuses on full-text search quick starts; upstash-workflow-js supports defining and managing durable multi-step workflows; upstash-box-js and upstash-box-py provide sandboxed cloud containers with shell, filesystem, and git; upstash-cli lets you drive the Upstash Developer API from the terminal; and upstash-redis-start provisions a zero-config, no-signup scratch Redis database for agents.
How do I install the Upstash Agent Skills DeepSeek plugin?
To install this plugin in DeepSeek Harness, ensure pnpm is on your PATH, then run:
dsh plugin --profile web add github:upstash/skills
After installation, start the harness:
dsh web
Store your Upstash credentials inside a session using /upstash-login YOUR_EMAIL YOUR_API_KEY. Credentials are saved to ~/.dsh/.credentials.yaml, and the MCP server connects as soon as both are set. The skills work without credentials.
How do I configure the MCP server for the Upstash Agent Skills DSH-Plugin?
The plugin includes the Upstash MCP server, so no extra configuration is needed within DeepSeek Harness. For other platforms like Claude Code or Cursor, refer to the README examples, such as running claude mcp add upstash -- npx -y @upstash/mcp-server@latest --email YOUR_EMAIL --api-key YOUR_API_KEY or adding the corresponding configuration to .cursor/mcp.json.
Which platforms are compatible with the Upstash Agent Skills DSH plugin?
Beyond DeepSeek Harness, this plugin is compatible with Claude Code, Cursor, OpenAI Codex, and the Agent Skills standard. It can be installed via each platform's plugin marketplace or CLI, for example using /plugin marketplace add upstash/skills and /plugin install upstash@upstash in Claude Code, or codex plugin marketplace add upstash/skills and codex plugin add upstash@upstash in OpenAI Codex.
How can I contribute to the Upstash Agent Skills DeepSeek Harness plugin?
To update or add skills, edit the files in the corresponding skill folder (e.g., skills/upstash-qstash-js/) and run npm run build to regenerate the build artifacts. The repository is maintained by Upstash, last updated in August 2026, with 19 stars and 5 forks, and welcomes community contributions.