Codex 完整教程：从安装到完成第一个小项目

npm install -g @openai/codex@latest
codex --version
codex --help

codex login
codex login status

printenv OPENAI_API_KEY | codex login --with-api-key

mkdir codex-catalog-lab
cd codex-catalog-lab
git init
python3 --version
git --version
pwd

"""Small catalogue validator. The starter has one intentional type bug."""
import argparse
import json
import sys
from pathlib import Path


def is_count(value):
    # Exercise bug: bool is a subclass of int in Python.
    return isinstance(value, int) and value >= 0


def validate(products):
    if not isinstance(products, list):
        raise ValueError("top level must be a list")
    issues = []
    for row, product in enumerate(products, start=1):
        if not isinstance(product, dict):
            issues.append({"row": row, "field": "*", "code": "not_object"})
            continue
        sku = product.get("sku")
        if not isinstance(sku, str) or not sku.strip():
            issues.append({"row": row, "field": "sku", "code": "required"})
        for field in ("price_cents", "stock"):
            if not is_count(product.get(field)):
                issues.append({
                    "row": row,
                    "field": field,
                    "code": "non_negative_integer_required",
                })
    return issues


def main():
    parser = argparse.ArgumentParser()
    parser.add_argument("input", type=Path)
    args = parser.parse_args()
    try:
        products = json.loads(args.input.read_text(encoding="utf-8"))
        issues = validate(products)
    except (OSError, ValueError) as exc:
        print(f"Input error: {exc}", file=sys.stderr)
        return 2
    print(json.dumps({"checked": len(products), "issues": issues}, indent=2))
    return 1 if issues else 0


if __name__ == "__main__":
    raise SystemExit(main())

[
  {"sku": "NOTEBOOK", "price_cents": 1200, "stock": 4},
  {"sku": "MUG", "price_cents": true, "stock": 0}
]

import json
import subprocess
import sys
import tempfile
import unittest
from pathlib import Path

from catalog import validate


def item(**changes):
    return {"sku": "MUG", "price_cents": 1200, "stock": 3, **changes}


def count_issue(field):
    return [{
        "row": 1,
        "field": field,
        "code": "non_negative_integer_required",
    }]


class CatalogueTests(unittest.TestCase):
    def test_valid(self):
        self.assertEqual(validate([item()]), [])

    def test_zero_is_valid(self):
        self.assertEqual(validate([item(price_cents=0, stock=0)]), [])

    def test_negative_price(self):
        self.assertEqual(validate([item(price_cents=-1)]),
                         count_issue("price_cents"))

    def test_negative_stock(self):
        self.assertEqual(validate([item(stock=-1)]), count_issue("stock"))

    def test_boolean_price(self):
        self.assertEqual(validate([item(price_cents=True)]),
                         count_issue("price_cents"))

    def test_boolean_stock(self):
        self.assertEqual(validate([item(stock=False)]), count_issue("stock"))

    def test_float_price(self):
        self.assertEqual(validate([item(price_cents=12.5)]),
                         count_issue("price_cents"))

    def test_string_stock(self):
        self.assertEqual(validate([item(stock="3")]), count_issue("stock"))

    def test_missing_stock(self):
        product = item()
        del product["stock"]
        self.assertEqual(validate([product]), count_issue("stock"))

    def test_blank_sku(self):
        self.assertEqual(validate([item(sku=" ")]),
                         [{"row": 1, "field": "sku", "code": "required"}])

    def test_non_object(self):
        self.assertEqual(validate([42]),
                         [{"row": 1, "field": "*", "code": "not_object"}])

    def test_top_level(self):
        with self.assertRaises(ValueError):
            validate({"sku": "MUG"})

    def run_cli(self, text):
        with tempfile.TemporaryDirectory() as directory:
            path = Path(directory) / "input.json"
            path.write_text(text, encoding="utf-8")
            return subprocess.run(
                [sys.executable, "-B", "catalog.py", str(path)],
                text=True, capture_output=True, check=False,
            )

    def test_cli_valid(self):
        result = self.run_cli(json.dumps([item()]))
        self.assertEqual(result.returncode, 0)
        self.assertEqual(json.loads(result.stdout),
                         {"checked": 1, "issues": []})

    def test_cli_finding(self):
        result = self.run_cli(json.dumps([item(stock=-1)]))
        self.assertEqual(result.returncode, 1)
        self.assertEqual(json.loads(result.stdout)["issues"],
                         count_issue("stock"))

    def test_cli_bad_json(self):
        result = self.run_cli("{")
        self.assertEqual(result.returncode, 2)
        self.assertIn("Input error:", result.stderr)


if __name__ == "__main__":
    unittest.main()

__pycache__/
*.pyc
runs/

# Catalogue practice

Use Python 3 standard library only. Work from the repository root.
The input is a JSON list. Each row is an object with a non-empty sku.
price_cents and stock must be non-negative integers; booleans are invalid.
Zero is valid. The CLI exits 0 for clean data, 1 for findings, 2 for bad input.

Run: python3 -B -m unittest -v
For the type-bug task, edit catalog.py only. Keep tests and fixtures unchanged.
Do not install dependencies or use the network for this exercise.
Do not commit, push, or deploy. Report actual commands and exit codes.

python3 -B -m unittest -v
printf "tests exit=%s\n" "$?"
python3 -B catalog.py products.json
printf "catalog exit=%s\n" "$?"

git add catalog.py test_catalog.py products.json AGENTS.md .gitignore
git status --short
git diff --check
git diff -- catalog.py

codex --sandbox read-only --ask-for-approval never

只读检查这个练习项目，不改文件，不运行测试，不联网。
阅读 AGENTS.md、catalog.py、test_catalog.py 和 products.json。
找出为什么布尔值价格没有被拒绝，给出文件、函数和条件表达式。
说明零值应该如何处理，并列出你尚未验证的部分。

先给修复计划，暂不修改。
范围只含 catalog.py 的整数判断，保留零值合法、负数非法。
保持测试与 products.json 不变，说明怎样验证没有扩大改动。

codex --sandbox workspace-write --ask-for-approval on-request

修复 catalog.py：price_cents 与 stock 必须是非负整数，拒绝 bool。
只改 catalog.py，保留零值合法，不改测试、样本和项目规则。
先读 AGENTS.md；若你发现需要扩大范围，停下来解释。
运行 python3 -B -m unittest -v，报告实际输出与退出码。
不要安装依赖、联网、提交、推送或部署。

git status --short
git diff --name-only
git diff --check
git diff -- catalog.py test_catalog.py products.json AGENTS.md
python3 -B -m unittest -v
printf "tests exit=%s\n" "$?"
python3 -B catalog.py products.json
printf "catalog exit=%s\n" "$?"

{
  "checked": 2,
  "issues": [
    {
      "row": 2,
      "field": "price_cents",
      "code": "non_negative_integer_required"
    }
  ]
}

def is_count(value):
    return type(value) is int and value >= 0

patch_file=$(mktemp ../catalog-repair.XXXXXX)
git diff -- catalog.py > "$patch_file"
printf "Saved: %s\n" "$patch_file"
git restore --worktree -- catalog.py
git diff -- catalog.py
python3 -B -m unittest -v

---
name: catalog-check
description: Review this catalogue validator after a local code change.
---

1. Read AGENTS.md and the current diff. Do not change files.
2. Run python3 -B -m unittest -v from the repository root if permitted.
3. Check that zero is accepted and booleans are rejected.
4. Report changed files, test exit code, and any unverified behavior.
5. Do not commit, push, deploy, or install packages.

codex mcp add openaiDeveloperDocs \
  --url https://developers.openai.com/mcp
codex mcp list

只读审查 catalog.py、test_catalog.py 与 products.json。
检查整数规则是否拒绝布尔值、是否保留零值。
不要编辑文件，不要运行测试，不要联网或使用外部工具。
报告具体文件、函数、条件表达式和仍需人工验证的事项。

#!/usr/bin/env bash
set -eu
mkdir -p runs
run_dir=$(mktemp -d runs/check.XXXXXX)
if codex -a never exec --sandbox read-only --color never --json \
    -o "$run_dir/final.txt" - < check.prompt.txt \
    > "$run_dir/events.jsonl" 2> "$run_dir/stderr.log"; then
  if [ ! -s "$run_dir/final.txt" ]; then
    printf 'Missing final message: %s\n' "$run_dir" >&2
    exit 2
  fi
else
  status=$?
  printf 'Codex exit %s; logs: %s\n' "$status" "$run_dir" >&2
  exit "$status"
fi
python3 -B -m unittest -v
printf 'Review files: %s\n' "$run_dir"