Skip to content

Instantly share code, notes, and snippets.

View MajorTal's full-sized avatar

Tal Weiss MajorTal

View GitHub Profile
@MajorTal
MajorTal / README.md
Created June 23, 2026 17:33
How does your AI coding agent talk to you? Mine Claude Code + Codex sessions for each side's signature expressions, using sentence embeddings.

How does your AI coding agent talk to you?

A single Python script that reads all your local Claude Code and Codex sessions and extracts the signature expressions of each side of the conversation: how Claude talks, how Codex talks, and how you talk to them.

It uses sentence embeddings, so it groups meaning, not just exact words: "let me check the file", "now let me verify the file", and "let me confirm the file" all collapse into one recurring expression.

@MajorTal
MajorTal / bad.eml
Created May 28, 2026 18:56
phishing gmail
Delivered-To: major.tal@gmail.com
Received: by 2002:a05:622a:6208:b0:50e:3693:c697 with SMTP id hj8csp48723qtb; Wed, 27 May 2026 15:54:03 -0700 (PDT)
X-Received: by 2002:a05:7022:2528:b0:135:a6a:fcb9 with SMTP id a92af1059eb24-1365fb5a001mr8987119c88.28.1779922443599; Wed, 27 May 2026 15:54:03 -0700 (PDT)
ARC-Seal: i=1; a=rsa-sha256; t=1779922443; cv=none; d=google.com; s=arc-20240605; b=Epf5bt6QNI/A1//8PWkVZhebAkQjy91G0TrwSv0Tx4L45pY7lv42dBqJeOt0nmpDm/ v+tHYBsND/0sLZ53vW5sSQtIcXyp31zkXm3Q4In+urJmtfiNhASx9ONl/UwMbJWXJ6Hb VXwko8Z1FetI9c15oj7/+T29XOhM6jlR026ao7Ibte1/FK9PGr418sCTdlFyYvwGAnOV xd8/ymro1vgimhuTPYRvjMKIVQYqbPFtrbWeZ9iK6YxfuwjGP0zeC5p/rQXGWIU9W7jm Ei3O5FpV/AexsDmueTkgrPhDNR+Xpei6Km+fNzerzLGyaMDTaVNAz6nIHTnw36mWZFes 6bng==
ARC-Message-Signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=google.com; s=arc-20240605; h=to:from:subject:message-id:gmsai:feedback-id:date:mime-version :dkim-signature; bh=76A64n+rBRzraJefgEc8fnISCxHn0j9nPtMzGJl3SXI=; fh=h1MwstUJpdrjbyv4544rkNMZhKrRppL1UTmFRWhIqw4=; b=GaysoqhmCl
@MajorTal
MajorTal / SKILL.md
Created February 26, 2026 09:27
Claude Code /consult skill — send project context to GPT-5.2 Pro for external consultation
name consult
description Send project context and a question to GPT-5.2 Pro for an external consultation. Gathers docs, code, and context automatically, submits via OpenAI Responses API in background mode, and notifies when the response arrives. Use /consult-results to read the response.
user-invocable true
disable-model-invocation true
allowed-tools Read, Glob, Grep, Bash(uv run *)
argument-hint <your question or request for the consultant>

/consult — External Consultation via GPT-5.2 Pro

from unittest.mock import Mock
import pytest
from livekit.agents import Agent, AgentSession
from livekit.plugins import openai
class MockEleanor(Agent):
"""Mock Eleanor agent that uses a regular LLM for testing."""
def __init__(self, context: dict) -> None:
local ReplicatedStorage = game:GetService("ReplicatedStorage")
local ServerStorage = game:GetService("ServerStorage")
local ShopItems = require(ServerStorage.ShopItems)
local buyItemEvent = ReplicatedStorage:WaitForChild("BuyItemEvent")
buyItemEvent.OnServerEvent:Connect(function(player, itemType)
local itemDef = ShopItems[itemType]
if not itemDef then
warn("Unknown item:", itemType)
local tool = script.Parent
local event = game.ReplicatedStorage:WaitForChild("GlobalPlaceRequest")
local itemType = tool:WaitForChild("ItemType").Value
local player = game.Players.LocalPlayer
local mouse = player:GetMouse()
tool.Activated:Connect(function()
if mouse.Target then
local position = mouse.Hit.p
event:FireServer(itemType, position)
local ReplicatedStorage = game:GetService("ReplicatedStorage")
local ServerStorage = game:GetService("ServerStorage")
local ShopItems = require(ServerStorage.ShopItems)
local function snapToGrid(position, gridSize)
return Vector3.new(
math.floor(position.X / gridSize + 0.5) * gridSize,
position.Y,
math.floor(position.Z / gridSize + 0.5) * gridSize
)
local function savePoles(player)
local roomName = player:GetAttribute("Room")
local room = workspace.PlayerRooms:FindFirstChild(roomName)
if not room then return end
local polesData = {
poles = {},
roomCenter = {
x = room.Position.X,
y = room.Position.Y,
local function loadPoles(player)
local previousDataSuccess, polesData = pcall(function()
return poleDataStore:GetAsync(player.UserId)
end)
if not previousDataSuccess or not polesData then
warn("No previous poles data found for player:", player.Name)
return
end
local DataStoreService = game:GetService("DataStoreService")
local poleDataStore = DataStoreService:GetDataStore("PolePlacementData")
local Players = game:GetService("Players")
local GRID_SIZE = 4 -- Same grid size you used earlier
-- Helper to snap positions
local function snapToGrid(position)
return Vector3.new(