ms-agent

name	ms-agent
description	Access ms-agent's advanced AI capabilities via MCP tools: deep research with async submit/check/get pattern, web search (arxiv/exa/serpapi), LSP code validation (TypeScript/Python/Java), concurrent-safe file editing, and agent delegation for complex multi-step tasks. Use when the user asks to research a topic, search the web, validate code, edit files precisely, or delegate a complex task to a specialized agent. Requires ms-agent (pip install ms-agent).
metadata	{"nanobot":{"emoji":"🤖","requires":{"bins":["python3"],"env":[]}}}

ms-agent Skills

This skill connects you to the ms-agent Capability Gateway — a unified interface to ms-agent's projects, components, and atomic tools, exposed as MCP tools.

Setup

Verify ms-agent is installed:

python scripts/check_ms_agent.py

The MCP server must be configured in your agent's config. See below for nanobot-specific setup.

nanobot config.json

{
  "tools": {
    "mcpServers": {
      "ms-agent": {
        "command": "python3",
        "args": ["-m", "ms_agent.capabilities.mcp_server"],
        "env": {"MS_AGENT_OUTPUT_DIR": "/path/to/workspace"},
        "toolTimeout": 120,
        "enabledTools": ["*"]
      }
    }
  }
}

Once configured, all 14 capabilities below are available as MCP tools (prefixed mcp_ms-agent_<tool_name> in nanobot).

Capability Index

Research & Search

Tool	Type	Duration	Reference
submit_research_task	async submit	seconds	deep-research.md
check_research_progress	async poll	seconds	deep-research.md
get_research_report	async result	seconds	deep-research.md
deep_research	sync (blocks)	hours	deep-research.md
web_search	instant	seconds	web-search.md

Agent Delegation

Tool	Type	Duration	Reference
delegate_task	sync (blocks)	minutes	agent-delegate.md
submit_agent_task	async submit	seconds	agent-delegate.md
check_agent_task	async poll	seconds	agent-delegate.md
get_agent_result	async result	seconds	agent-delegate.md
cancel_agent_task	async cancel	seconds	agent-delegate.md

The tools parameter for agent delegation currently supports the basic tool components web_search, file_system, and todo_list (filesystem is kept as a backward-compatible alias).

Code Validation

Tool	Type	Duration	Reference
lsp_check_directory	full scan	minutes	lsp-code-server.md
lsp_update_and_check	incremental	seconds	lsp-code-server.md

File Editing

Tool	Type	Duration	Reference
replace_file_contents	content-match	seconds	filesystem-tools.md
replace_file_lines	line-range	seconds	filesystem-tools.md

Quick Decision Guide

User wants to...
│
├── Research a topic in depth (20-60 min)
│   └── submit_research_task → check_research_progress → get_research_report
│
├── Search the web for quick info
│   └── web_search(query="...", engine_type="arxiv|exa|serpapi")
│
├── Delegate a complex multi-step task to an AI agent
│   ├── Short task (< 3 min) → delegate_task(query="...")
│   └── Long task → submit_agent_task → check_agent_task → get_agent_result
│
├── Validate code for errors
│   ├── Full project → lsp_check_directory(directory="src/", language="typescript")
│   └── Single file → lsp_update_and_check(file_path="...", content="...", language="...")
│
└── Edit a file precisely
    ├── Know exact text → replace_file_contents (concurrent-safe)
    └── Know line numbers → replace_file_lines

Async Pattern

Several capabilities (deep research, agent delegation) support an async submit/check/get pattern. This does NOT block the agent — continue handling other messages while the task runs in the background.

1. submit_*_task(...)      → returns task_id immediately
2. check_*_task(task_id)   → poll status (repeat every few minutes)
3. get_*_result/report(task_id)   → retrieve final result when completed
4. cancel_*_task(task_id)  → cancel if no longer needed

Read the reference files for detailed SOP workflows for each capability.