Skip to main content

reactive-resume/packages/dsh-plugin

41.3kStars4.6kForks108Issues126Watchers

Integrate Reactive Resume's MCP tools with DeepSeek Harness, enabling AI to directly read, modify, and lock users' online resumes and job applications.

Evidence5/5methodologySourceInstallMaintenanceDSH versionSecurity scan
Machine-auditedInstall commandRepo verifieddsh-plugin topicLicenseREADMEAI wiki
Language
TypeScript
License
MIT
Branch
main
agent-skillsaidsh-pluginhacktoberfestjavascriptmcp-serverreactresume-builder

Install

cmdweb profile
$ dsh plugin --profile web add dsh-plugin-reactive-resume

Run the command above in your terminal to install this plugin via the dsh CLI. You can switch Profile in the top-right corner. New to dsh? Read the beginner tutorial

Install via your agent

Install the DeepSeek Harness plugin amruthpillai/reactive-resume/packages/dsh-plugin for me: review the repository at https://github.com/amruthpillai/reactive-resume first, then run the install command and verify the plugin loads successfully.

Paste this instruction to the DSH Web GUI assistant — it will install and verify for you.

One-Line Positioning

Integrate the MCP toolset of the online resume service Reactive Resume into DeepSeek Harness, enabling AI to read, modify, lock, and track user resumes and job applications like operating files. Includes a system prompt tutorial that teaches the model how to correctly write resume data, avoiding common patch path errors.

Core Capabilities

  • Bridge Reactive Resume's MCP endpoint via streamable-http, exposing 28+ resume and job application tools to the model
  • Inject a dedicated system prompt section that standardizes the model's resume editing process (list IDs first, read by ID, RFC 6902 patch, UUID-keyed entries, lock state handling)
  • Read and forward user API Key, defaulting to the official hosted address https://rxresu.me, also supports self-hosted instances
  • When API Key is not configured, only print warning without mounting tools, ensuring profile can still start
  • Customize tool namespace at the plugin layer (serverName) to avoid tool name conflicts when multiple Reactive Resume instances are connected

Technical Implementation

  • Language: TypeScript (ESM, TypeScript 7)
  • Key Dependencies: @deepseek-ai/cordis (plugin host framework), @deepseek-ai/dsh-mcp-client (MCP bridging), @deepseek-ai/dsh-system-prompt (system prompt injection), @deepseek-ai/schemastery (config schema validation)
  • Architecture Pattern: Cordis bundle patch declarative injection — cordis.patch.yml inserts a line reactive-resume at installation, the plugin apply(ctx, config) hook calls ctx.plugin() to mount the MCP client when Harness starts, and appends prompts via ctx.systemPrompt.section()
  • Entry File: src/index.ts (apply function + Config schema + name/inject exports)

Use Cases

Users already managing online resumes with Reactive Resume who want DeepSeek Harness conversations to directly read, modify, batch update resumes or job applications. The pain point is: manually editing resume field names and copying work experience descriptions is tedious, and letting the model directly operate on data risks UUID guessing errors and lock state pitfalls — this plugin delivers both tools and operation specifications to the model.

Prerequisites and Compatibility

DependencyMinimum VersionDescription
Node.js^22.19.0 || >=24.0.0From package.json#engines
@deepseek-ai/cordis^4.0.1Host framework, peerDependency
@deepseek-ai/dsh-mcp-client^0.1.0-rc.6MCP bridge client
@deepseek-ai/dsh-system-prompt^0.1.0-rc.6System prompt injection interface
@deepseek-ai/schemastery^3.18.1Config schema validation
PlatformCross-platformNo native module dependencies
Native modulesNonePure JS implementation

Installation

dsh plugin --profile web add github:amruthpillai/reactive-resume/packages/dsh-plugin

Configuration Options

ConfigTypeDescriptionDefault
apiKeystringAPI Key from <url>/dashboard/settings/api-keys; empty string means do not mount tools, just print warning''
urlstringRoot address of Reactive Resume instance; change to your own address for self-hostinghttps://rxresu.me
serverNamestringTool namespace prefix, tools will appear as mcp__<serverName>__<toolName> on the model side; only letters, numbers, underscores, hyphens allowed, max 32 charactersresume
toolCallTimeoutMsnumberSingle tool call timeout (ms)60000

FAQ

Q: Do I need to configure apiKey immediately after installation?

A: No. When the apply function detects an empty apiKey, it only prints a warning and returns directly without calling ctx.plugin(). So without configuration, the profile can still start normally, and the model won't see Reactive Resume tools.

Q: Can I self-host Reactive Resume and connect it?

A: Yes. Change the url configuration to your self-hosted instance address (e.g., http://localhost:3000), the plugin will automatically strip trailing slashes and append the /mcp endpoint. Remember to also generate an API Key on the self-hosted instance.

Q: Can I expose only some tools instead of all for AI use?

A: No. Harness's ctx.tools.restrict() can only be called in agent context, while the plugin apply receives plugin-level context, so toolset trimming is not currently possible — all tools published by Reactive Resume will be exposed to the model.

Q: What is the serverName field for? What happens if I change it?

A: It's the tool namespace prefix; the actual tool name the model sees is mcp__<serverName>__<rawName>. The default is resume. If you want to connect both official and self-hosted instances simultaneously, change their serverName to different values (e.g., resume-hosted / resume-self) to avoid name conflicts.

Q: What does the plugin write to the system prompt?

A: It appends a "Reactive Resume" section via ctx.systemPrompt.section() covering: use list_resumes before edit_resume first, then read_resume, use RFC 6902 JSON Patch instead of overwrite, section entries are object arrays with UUID ids, locked resumes must be unlock_resume first, deletion actions require explicit user authorization, etc.

Q: What happens when a tool call times out?

A: Default single call timeout is 60 seconds (toolCallTimeoutMs: 60000), timeout is handled by dsh-mcp-client. If your network or resumes are particularly large, you can increase this value in your profile's cordis.patch.yml.

Difficulty Level

Beginner — only one required configuration (apiKey), all others have defaults. Regular users can use it once they get an API Key.

Known Issues and Limitations

  • Toolset cannot be trimmed at the plugin layer; all tools published by Reactive Resume will be exposed to the model (src/index.ts:23-30)
  • When apiKey is not configured, only warn without mounting, but there's no switch to disable warning messages (src/index.ts:44)
  • serverName has a 32-character limit and only allows [A-Za-z0-9_-]; exceeding or containing illegal characters will throw errors during schema validation (src/config.ts:4 / src/config.ts:28-32)
  • Plugin directly uses failOnStartupError: true so MCP startup failures bubble up to the host; network fluctuations during startup can cause profile startup failures (src/index.ts:56)

Read the usage guide →

Install steps, key points, FAQ and compatibility for this plugin — auto-derived from indexed fields.

Listing badge

Listed on deepseek-plugin.org
[![Listed on deepseek-plugin.org](https://img.shields.io/badge/listed_on-deepseek--plugin.org-007EC6)](https://deepseek-plugin.org/plugins/amruthpillai/reactive-resume/packages/dsh-plugin)

Paste this markdown into your GitHub README to link back to this listing. The badge only states the listing — not a security endorsement.

← Back to plugin directory