Introduction

The core philosophy of DeepSeek Harness (DSH) is “everything is a plugin.” As the plugin list grows, developers often face two concrete questions: Are installed plugins actually loaded and running? How can they find public evaluation boards for benchmarking? @trapstreet/dsh-trapstreet aims to solve both problems.

This is a network-connected tool maintained by developer trapstreet under the MIT open source license. It does not require you to manually troubleshoot configuration; instead, it directly reads the internal state of the Harness runtime and connects to public evaluation data.

Installation and Dependencies

Before installing, ensure your environment satisfies the following dependencies:

  • @deepseek-ai/cordis
  • @deepseek-ai/dsh-system-prompt
  • @deepseek-ai/dsh-tools

The installation command is as follows:

dsh plugin --profile <name> add @trapstreet/dsh-trapstreet

You can also skip npm and add it directly from GitHub:

dsh plugin --profile <name> add github:trapstreet/dsh-trapstreet

Note: The installation directory path may cause dependency resolution failures, so it is recommended to install using a tarball package.

Core Features

The plugin provides two main tools and one classification capability.

1. trapstreet_checkup: Local Plugin Status Check

trapstreet_checkup is a network-free tool that does not require an API key. It directly reads the loader tree of the Harness runtime and lists installed plugin entries along with their activation status.

It can identify plugins that were “installed successfully but did not enter the loader tree” or whose “module loading failed in the loader tree.” By passing the classify parameter, it can also classify plugins based on a community catalog (such as skillhub.cn), identifying conversational or orchestration plugins.

Usage examples:

# 检查插件状态
trapstreet_checkup

# 检查状态并分类
trapstreet_checkup classify

2. trapstreet_boards: Public Evaluation Board Lookup

trapstreet_boards requires a network connection and is used to find public evaluation boards on trapstreet.run. This is a read-only operation and does not involve account login, API keys, or starting runs.

It lists the total number of evaluation boards and displays them by category: boards that include DeepSeek benchmarks, boards that have not tested DeepSeek, and empty boards. You can pass a board ID to obtain details, including what was tested, code versions, and which entries come from DeepSeek models.

Usage examples:

trapstreet_boards

Configuration

The plugin supports the following environment variable settings:

  • TRAPSTREET_BASE_URL: The API base URL, defaulting to https://trapstreet.run.
  • TRAPSTREET_CALL_LOG: If this path is set, each tool call appends one log line to the file, making it easier to verify whether the model called the tool correctly.

Notes

  • Read-only data: trapstreet_boards only fetches public data from trapstreet.run; it does not start any runs or write local files.
  • Catalog relationship: Community catalogs (such as skillhub.cn) have no official affiliation with DeepSeek or Harness itself, and serve only as resource indexes.
  • Installation method: As mentioned earlier, the installation directory path may cause dependency resolution failures, so a tarball-based installation must be used.

Summary

@trapstreet/dsh-trapstreet provides a complete workflow from local plugin health checks to remote evaluation board queries. For developers who need to validate their environment or find benchmark data, this is a practical helper tool.