Loading ToolBench

Reading the current corpus…

LLM-RAG-Agent-Tutorial MCP server · F · ToolBench

LLM-RAG-Agent-Tutorial

github.com/mac999/LLM-RAG-Agent-Tutorial

Tutorial repository demonstrating LLM, RAG (Retrieval-Augmented Generation), and Agent implementations with various transformer models, fine-tuning techniques, and database integrations

2026-07-28+2 tools

21/100

Definition Quality28

Protocol Readiness5

SupportabilityNot measured-

Spec posture

Inferred effective spec 2026-07-28+·current baseline 2026-07-28

Static source inference · medium confidence

Current-spec patterns detected

Summary

This repository is a tutorial/educational project, not a production MCP server. No MCP protocol implementation is visible in the source code. The two tools mentioned (search_news, python_repl_tool) appear in a single Python file (5-1.agent/lang-call-tools-news-python.py) but no tool registration, schema definitions, or MCP server scaffolding is evident. The repository contains Jupyter notebooks, tutorial materials, and requirements.txt files for various ML/transformer/finetuning exercises, but no functional MCP server code. Tool definitions cannot be verified from the provided source excerpt.

Tools (2)

python_repl_toolirreversiblesource verified27/100

Use this to execute python code. If you want to see the output of a value, you should print it out with `print(...)`. This is visible to the user.

search_newsread onlysource verified28/100

Search naver.com news article and returns the body content.

Issues (8)

CRITICALtool ↗

No MCP server implementation detected. Repository is a tutorial/educational project with Python scripts, not a functioning MCP server.

search_newspython_repl_tool

CRITICALtool ↗

Tool schemas completely absent. No JSON Schema definitions visible for input parameters.

search_newspython_repl_tool

HIGHtool-description ↗

Tool descriptions are minimal and lack context about when/how to use them. 'Search naver.com news article and returns the body content.' (55 chars) lacks WHEN to use it, error handling, or dependencies. python_repl_tool description (121 chars) is generic template text.

search_newspython_repl_tool

CRITICALconfirmation-request ↗

python_repl_tool is an IRREVERSIBLE operation with unrestricted code execution. No confirmation step, no sandboxing, no output limits documented. Violates the confirmation-request pattern for dangerous operations.

Recommendations

Implement a proper MCP server using the official Python SDK (mcp package). This repository is educational material, not a production server.
Add complete JSON Schema definitions for all tool inputs with type, description, and constraints for each parameter.
Document output schemas explicitly. For search_news, specify: does it return {title, url, body, publish_date}? For python_repl_tool, specify: does it return {stdout, stderr, return_value}?
Expand search_news description to ~100-150 chars: 'Fetch and parse article content from a Naver News URL. Returns structured data including title, body, publication date, and author. Fails gracefully if URL is invalid or article is not found.'
For python_repl_tool, add: 'Execute Python code in a sandboxed REPL environment with a 30-second timeout and 1MB output limit. Logs all code executed. Use for data analysis, chart generation, or light computation only. Do not attempt file I/O, network calls, or system commands.'
Add explicit error handling documentation: 'search_news returns {error: string} if the URL is malformed, the domain is not naver.com, or the article does not exist. Example: {error: "Invalid URL format"}. If you receive an error, ask the user to provide a valid Naver News article link.'
For python_repl_tool, require a dry-run/preview mode or add a confirmation step before executing code that modifies state or has side effects.
Sanitize python_repl_tool inputs: block imports of dangerous modules (os, subprocess, socket, requests if not sandboxed), validate code for credential leaks, and enforce output size caps.

Score history

Overall score trend

↓ 3 points across a rubric change (v1 → v2)

21/100

Scored	Grade	Overall	Spec posture	Rubric
2026-09-22	F	21	2026-07-28+	v2
2026-03-09	F	24	-	v1