AtlSciData

ClaudeSDKBasedResearchAssistant — Development skill for Claude Code

Development community

a small CLI agent built on a real hand-written MCP server talking to a Claude Agent SDK client.

How to install ClaudeSDKBasedResearchAssistant

This entry records only its repository, not the path inside it, so there is no exact command to give. Open AtlSciData/ClaudeSDKBasedResearchAssistant and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What ClaudeSDKBasedResearchAssistant does

a small CLI agent built on a real hand-written MCP server talking to a Claude Agent SDK client. that answers questions about a company's business (segments, results, strategy) by retrieving grounded passages from public investor documents

Alternatives in Development

Iterate Implementation Plan — You are tasked with updating existing implementation plans based on user feedback 10k ★
Market Segments — Identify 3-5 customer segments with JTBD and product fit 7.8k ★
Notebooklm Skill — Use this skill to enable Claude Code to communicate directly with your Google NotebookLM notebooks 5.9k ★

README

AIG GenAI Research Assistant

A small, authenticated RAG agent for analyzing AIG's public investor communications (quarterly results presentations and investor day materials). Built to demonstrate a production-shaped agentic architecture: a real authentication flow, a tool-using agent over the Claude Agent SDK, and a retrieval layer exposed through the Model Context Protocol (MCP) rather than hand-rolled inside the agent.

Architecture

                     ┌──────────────────┐
   Browser  ───────► │  Frontend (SPA)  │
                     │  static HTML/JS  │
                     └────────┬─────────┘
                              │ HTTPS
                              ▼
                     ┌──────────────────────────────┐
                     │   FastAPI backend             │
                     │   backend/app/main.py         │
                     │                                │
                     │   POST /api/login ────────────┼───► Amazon Cognito
                     │     (USER_PASSWORD_AUTH)      │      (user pool)
                     │                                │
                     │   POST /api/chat               │
                     │     (Bearer JWT, verified      │◄──── JWKS
                     │      against Cognito JWKS)     │
                     │                                │
                     │   GET  /healthz                │
                     └────────┬───────────────────────┘
                              │ per-session
                              ▼
                     ┌──────────────────────────────┐
                     │  Claude Agent SDK session     │
                     │  backend/app/agent_session.py │
                     │  (one ClaudeSDKClient per      │
                     │   session_id, kept in memory) │──────► AWS Bedrock
                     └────────┬───────────────────────┘        (Claude model)
                              │ spawns as subprocess, talks stdio