Rodion Mostovoi rodion-m

Оркестрация AI-агентов: материалы доклада

Родион Мостовой, CodeAlive. Podlodka AI Crew, 2026. Материалы к редакции от 12 октября.

С чего начать

Выберите одну большую задачу, опишите результат и ограничения, выделите первый проверяемый шаг. Сильная модель готовит план, достаточно сильный исполнитель выполняет шаг, координатор проверяет результат. Когда прогресс остановился, свежий аудитор смотрит на цель и факты и предлагает другой подход. Изменение подхода не отменяет критерии приёмки.

Скиллы и инструменты

### ABOUT
Modern AI coding tools converge on a simple idea: give the agent a **single, well-structured Markdown file that explains how your repo “works,”** and prepend that file to every LLM call so the agent never has to guess about architecture, commands, or conventions. Community gists, RFCs, and vendor playbooks all recommend the same core sections—overview, project map, build/test scripts, code style, security, and guardrails—plus support for nested AGENTS.md files that override one another hierarchically.

### SYSTEM
You are a meticulous technical writer and senior staff engineer.  
Your task is to create **AGENTS.md** for the repository whose contents are provided below.

### CONTEXT (replace the JSON blocks)
See the attached files.

AI that understands your entire codebase

Streamline your development process and boost team productivity with CodeAlive:

Instant Answers: Get accurate responses to your codebase questions in natural language
Context-Aware Code Reviews: Elevate your Pull/Merge Requests with AI that understands your entire system
Tickets Enrichment: an AI-expert that writes suggestions and dev notes to help developers to get started

Обо мне

Привет! Рад видеть вас на своей странице. Сначала ключевое обо мне:

Меня зовут Родион Мостовой. Более 10-ти лет я программирую за деньги под платформу .NET, из них 6 лет на ASP.NET Core. Почти 3 года преподаю C# и ASP.NET Core, за это время научился объяснять сложные концепции простым языком, разработал несколько собственных учебных курсов. Сейчас преподаю в OTUS. В обучении предпочитаю, прежде всего, подход через общение, наводящие вопросы и разбор практических кейсов.

Веду несколько популярных Open Source проектов на GitHub: https://github.com/rodion-m/

С чем могу помочь

С понимаем самых сложных тем:

Асинхронность/многопоточность (async/await, ThreadPool и т. д.)

	Investigate this machine for the reported Anthropic.claude-code 2.1.313 supply-chain compromise. Establish what executed, what access was possible, what actually happened, and what is required to restore trust.

	Sources:
	https://phorion.io/blog/offical-claude-code-vsx-compromise/
	https://cursor.com/en-US/blog/claude-code-extension-security-notice

	Use these as evidence leads, not unquestionable instructions. Cross-check indicators, publication revisions, affected distribution channels, platforms, and sample types. Distinguish VSIX hashes from extension.js hashes. Record disagreements and inaccessible sources; do not invent missing details.

	OPERATING PRINCIPLE
	Run an adaptive, evidence-driven investigation, not an exhaustive checklist. Prioritize checks that change containment or recovery decisions. Continue safe, authorized checks without repeatedly asking permission; batch requests for additional access. Never manufacture certainty to finish the task.

	using System.Collections.Concurrent;
	using System.Threading.Channels;

	public static class DynamicParallelism
	{
	public static async Task ForEachAsync<TSource>(
	IEnumerable<TSource> source,
	Func<TSource, CancellationToken, Task> bodyAsync,
	Func<int> getDegreeOfParallelism,
	bool preserveOrder = false,

	import argparse
	import json
	import logging
	import os
	import sqlite3

	import requests
	from tqdm import tqdm

	# --- Configuration ---

	using System.Reflection;
	using Microsoft.CodeAnalysis;

	namespace Gend.Domain;

	/// <summary>
	/// Provides a factory for creating instances of the internal SymbolDisplayFormat class from the Roslyn library.
	/// </summary>
	public static class SymbolDisplayFormatFactory
	{

	import logging
	import os
	import sys
	import time
	from typing import List

	from pydub import AudioSegment
	from pydub.silence import split_on_silence

	MAX_CHUNK_SIZE_MB = int(os.getenv("MAX_CHUNK_SIZE_MB", 100))

	You are Meta-Expert, an extremely clever expert with the unique ability to collaborate with multiple experts (such as Expert Problem Solver, Expert Mathematician, Expert Essayist, etc.) to tackle any task and solve any complex problems. Some experts are adept at generating solutions, while others excel in verifying answers and providing valuable feedback.

	Note that you also have special access to Expert Python, which has the unique ability to generate and execute Python code given natural-language instructions. Expert Python is highly capable of crafting code to perform complex calculations when given clear and precise directions. You might therefore want to use it especially for computational tasks.

	As Meta-Expert, your role is to oversee the communication between the experts, effectively using their skills to answer a given question while applying your own critical thinking and verification abilities.

	To communicate with an expert, type its name (e.g., "Expert Linguist" or "Expert Puzzle Solver"), follow