Back to Details

codex-subagent

Compare original and translation side by side

🇺🇸

Original

English
🇨🇳

Translation

Chinese

Codex Subagent Skill

Codex子代理技能

Spawn autonomous subagents to offload context-heavy work. Subagents burn their own tokens, return only final results.
Golden Rule: If task + intermediate work would add 3,000+ tokens to parent context → use subagent.
生成自主子代理以卸载上下文密集型工作。子代理会消耗自身的tokens，仅返回最终结果。
**黄金准则：**如果任务+中间工作会给父代理上下文增加3000+ tokens → 使用子代理。

Intelligent Prompting

智能提示词设计

Critical: Parent agent must provide subagent with essential context for success.
关键：父代理必须为子代理提供成功所需的必要上下文。

Good Prompting Principles

优秀提示词原则

  1. Include relevant context - Give the subagent thorough context
  2. Be specific - Clear constraints, requirements, output format
  3. Provide direction - Where to look, what sources to prioritize
  4. Define success - What constitutes a complete answer
  1. 包含相关上下文 - 为子代理提供完整的上下文信息
  2. 明确具体 - 清晰的约束条件、要求、输出格式
  3. 提供方向 - 指定查找位置、优先使用的数据源
  4. 定义成功标准 - 明确什么是完整的答案

Examples

示例

❌ Bad: "Research authentication"
✅ Good: "Research authentication in this Next.js codebase. Focus on: 1) Session management strategy (JWT vs session cookies), 2) Auth provider integration (NextAuth, Clerk, etc), 3) Protected route patterns. Check /app, /lib/auth, and middleware files. Return architecture summary with code examples."
❌ Bad: "Search for Codex SDK"
✅ Good: "Find the most recent Codex SDK documentation and summarize key updates. Focus on: 1) Installation/quickstart, 2) Core API methods and parameters, 3) Breaking changes or deprecations. Prioritize official OpenAI docs and release notes. Return a concise summary with citations."
❌ Bad: "Find API endpoints"
✅ Good: "Find all REST API endpoints in this Express.js app. Look in /routes, /api, and /controllers directories. For each endpoint document: method (GET/POST/etc), path, auth requirements, request/response schemas. Return as markdown table."
❌ 错误示例： "研究认证"
✅ 正确示例： "研究此Next.js代码库中的认证机制。重点关注：1) 会话管理策略（JWT vs 会话Cookie），2) 认证提供商集成（NextAuth、Clerk等），3) 受保护路由模式。检查/app、/lib/auth和middleware文件。返回包含代码示例的架构总结。"
❌ 错误示例： "搜索Codex SDK"
✅ 正确示例： "查找最新的Codex SDK文档并总结关键更新。重点关注：1) 安装/快速开始指南，2) 核心API方法及参数，3) 破坏性变更或弃用内容。优先参考OpenAI官方文档和发布说明。返回带引用的简洁总结。"
❌ 错误示例： "查找API端点"
✅ 正确示例： "查找此Express.js应用中的所有REST API端点。查看/routes、/api和/controllers目录。为每个端点记录：请求方法（GET/POST等）、路径、认证要求、请求/响应 schema。以Markdown表格形式返回。"

Prompting Template

提示词模板

[TASK CONTEXT]
You are researching/analyzing [SPECIFIC TOPIC] in [LOCATION/CODEBASE/DOMAIN].

[OBJECTIVES]
Your goals:
1. [1st objective with specifics]
2. [2nd objective]
3. [3rd objective if needed]

[CONSTRAINTS]
- Focus on: [specific areas/files/sources]
- Prioritize: [what matters most]
- Ignore: [what to skip]

[OUTPUT FORMAT]
Return: [exactly what format parent needs]

[SUCCESS CRITERIA]
Complete when: [specific conditions met]
[任务上下文]
你正在[位置/代码库/领域]中研究/分析[具体主题]。

[目标]
你的目标：
1. [带具体细节的第一个目标]
2. [第二个目标]
3. [如有需要，第三个目标]

[约束条件]
- 重点关注：[具体区域/文件/数据源]
- 优先处理：[最重要的内容]
- 忽略：[需要跳过的内容]

[输出格式]
返回：[父代理所需的具体格式]

[成功标准]
完成条件：[具体满足的条件]

Model Selection

模型选择

Use Mini Model (gpt-5.1-codex-mini + medium)

使用迷你模型（gpt-5.1-codex-mini + medium）

Pure search only - no additional work after gathering info.
仅纯搜索 - 收集信息后无需额外工作。

Bash (Linux/macOS)

Bash（Linux/macOS）

bash
codex exec --dangerously-bypass-approvals-and-sandbox --skip-git-repo-check \
  -m gpt-5.1-codex-mini -c 'model_reasoning_effort="medium"' \
  "Search web for [TOPIC] and summarize findings"
bash
codex exec --dangerously-bypass-approvals-and-sandbox --skip-git-repo-check \
  -m gpt-5.1-codex-mini -c 'model_reasoning_effort="medium"' \
  "Search web for [TOPIC] and summarize findings"

PowerShell (Windows)

PowerShell（Windows）

powershell
codex exec --dangerously-bypass-approvals-and-sandbox --skip-git-repo-check `
  -m gpt-5.1-codex-mini -c 'model_reasoning_effort="medium"' `
  "Search web for [TOPIC] and summarize findings"
powershell
codex exec --dangerously-bypass-approvals-and-sandbox --skip-git-repo-check `
  -m gpt-5.1-codex-mini -c 'model_reasoning_effort="medium"' `
  "Search web for [TOPIC] and summarize findings"

Inherit Parent Model + Reasoning

继承父代理模型+推理能力

Multi-step workflows - search + analyze/refactor/generate:
多步骤工作流 - 搜索+分析/重构/生成：

Bash (Linux/macOS)

Bash（Linux/macOS）

bash
codex exec --dangerously-bypass-approvals-and-sandbox --skip-git-repo-check \
  -m "$MODEL" -c "model_reasoning_effort=\"$REASONING\"" \
  "Find auth files THEN analyze security patterns and propose improvements"
bash
codex exec --dangerously-bypass-approvals-and-sandbox --skip-git-repo-check \
  -m "$MODEL" -c "model_reasoning_effort=\"$REASONING\"" \
  "Find auth files THEN analyze security patterns and propose improvements"

PowerShell (Windows)

PowerShell（Windows）

powershell
codex exec --dangerously-bypass-approvals-and-sandbox --skip-git-repo-check `
  -m $MODEL -c "model_reasoning_effort=`"$REASONING`"" `
  "Find auth files THEN analyze security patterns and propose improvements"
powershell
codex exec --dangerously-bypass-approvals-and-sandbox --skip-git-repo-check `
  -m $MODEL -c "model_reasoning_effort=`"$REASONING`"" `
  "Find auth files THEN analyze security patterns and propose improvements"

Decision Logic

决策逻辑

Is task PURELY search/gather?
├─ YES: Any work after gathering?
│  ├─ NO → mini model
│  └─ YES → inherit parent
└─ NO → inherit parent
任务是否仅为搜索/收集？
├─ 是：收集后是否有额外工作？
│  ├─ 否 → 使用迷你模型
│  └─ 是 → 继承父代理模型
└─ 否 → 继承父代理模型

Basic Usage

基础用法

Bash (Linux/macOS)

Bash（Linux/macOS）

bash
undefined
bash
undefined

Get parent session settings (respects active profile; falls back to top-level)

获取父代理会话设置（遵循活跃配置文件；回退到顶级配置）

NOTE: codex-parent-settings.sh prints two lines; use mapfile to avoid empty REASONING.

注意：codex-parent-settings.sh输出两行；使用mapfile避免REASONING为空。

mapfile -t _settings < <(scripts/codex-parent-settings.sh) MODEL="${_settings[0]}" REASONING="${_settings[1]}"
mapfile -t _settings < <(scripts/codex-parent-settings.sh) MODEL="${_settings[0]}" REASONING="${_settings[1]}"

Spawn subagent (inherit parent)

生成子代理（继承父代理）

codex exec --dangerously-bypass-approvals-and-sandbox --skip-git-repo-check
-m "$MODEL" -c "model_reasoning_effort="$REASONING""
"DETAILED_PROMPT_WITH_CONTEXT"
codex exec --dangerously-bypass-approvals-and-sandbox --skip-git-repo-check
-m "$MODEL" -c "model_reasoning_effort="$REASONING""
"DETAILED_PROMPT_WITH_CONTEXT"

Safer prompt construction (no backticks / command substitution)

更安全的提示词构建（避免反引号/命令替换）

PROMPT=$(cat <<'EOF' [TASK CONTEXT] You are analyzing /path/to/repo.
[OBJECTIVES]
  1. Do X
  2. Do Y
[OUTPUT FORMAT] Return: path - purpose EOF ) codex exec --dangerously-bypass-approvals-and-sandbox --skip-git-repo-check
-m "$MODEL" -c "model_reasoning_effort="$REASONING""
"$PROMPT"
PROMPT=$(cat <<'EOF' [TASK CONTEXT] You are analyzing /path/to/repo.
[OBJECTIVES]
  1. Do X
  2. Do Y
[OUTPUT FORMAT] Return: path - purpose EOF ) codex exec --dangerously-bypass-approvals-and-sandbox --skip-git-repo-check
-m "$MODEL" -c "model_reasoning_effort="$REASONING""
"$PROMPT"

Pure search (use mini)

纯搜索（使用迷你模型）

codex exec --dangerously-bypass-approvals-and-sandbox --skip-git-repo-check
-m gpt-5.1-codex-mini -c 'model_reasoning_effort="medium"'
"SEARCH_ONLY_PROMPT"
codex exec --dangerously-bypass-approvals-and-sandbox --skip-git-repo-check
-m gpt-5.1-codex-mini -c 'model_reasoning_effort="medium"'
"SEARCH_ONLY_PROMPT"

JSON output for parsing

用于解析的JSON输出

codex exec --dangerously-bypass-approvals-and-sandbox --json "PROMPT" | jq -r 'select(.event=="turn.completed") | .content'
undefined
codex exec --dangerously-bypass-approvals-and-sandbox --json "PROMPT" | jq -r 'select(.event=="turn.completed") | .content'
undefined

PowerShell (Windows)

PowerShell（Windows）

powershell
undefined
powershell
undefined

Get parent session settings (respects active profile; falls back to top-level)

获取父代理会话设置（遵循活跃配置文件；回退到顶级配置）

$scriptPath = Join-Path $env:USERPROFILE ".codex\skills\codex-subagent\scripts\codex-parent-settings.ps1" $settings = & $scriptPath $MODEL = $settings[0] $REASONING = $settings[1]
$scriptPath = Join-Path $env:USERPROFILE ".codex\skills\codex-subagent\scripts\codex-parent-settings.ps1" $settings = & $scriptPath $MODEL = $settings[0] $REASONING = $settings[1]

Spawn subagent (inherit parent)

生成子代理（继承父代理）

codex exec --dangerously-bypass-approvals-and-sandbox --skip-git-repo-check 
   -m $MODEL -c "model_reasoning_effort=
"$REASONING
""
"DETAILED_PROMPT_WITH_CONTEXT"
codex exec --dangerously-bypass-approvals-and-sandbox --skip-git-repo-check 
   -m $MODEL -c "model_reasoning_effort=
"$REASONING
""
"DETAILED_PROMPT_WITH_CONTEXT"

Use here-string for multi-line prompts (avoids escaping issues)

使用here-string处理多行提示词（避免转义问题）

$PROMPT = @' [TASK CONTEXT] You are analyzing /path/to/repo.
[OBJECTIVES]
  1. Do X
  2. Do Y
[OUTPUT FORMAT] Return: path - purpose '@
codex exec --dangerously-bypass-approvals-and-sandbox --skip-git-repo-check 
   -m $MODEL -c "model_reasoning_effort=
"$REASONING
""
$PROMPT
$PROMPT = @' [TASK CONTEXT] You are analyzing /path/to/repo.
[OBJECTIVES]
  1. Do X
  2. Do Y
[OUTPUT FORMAT] Return: path - purpose '@
codex exec --dangerously-bypass-approvals-and-sandbox --skip-git-repo-check 
   -m $MODEL -c "model_reasoning_effort=
"$REASONING
""
$PROMPT

Pure search (use mini)

纯搜索（使用迷你模型）

codex exec --dangerously-bypass-approvals-and-sandbox --skip-git-repo-check 
  -m gpt-5.1-codex-mini -c 'model_reasoning_effort="medium"'
"SEARCH_ONLY_PROMPT"
codex exec --dangerously-bypass-approvals-and-sandbox --skip-git-repo-check 
  -m gpt-5.1-codex-mini -c 'model_reasoning_effort="medium"'
"SEARCH_ONLY_PROMPT"

Method 1 (Recommended): Use -o to output directly to file

方法1（推荐）：使用-o直接输出到文件

codex exec --dangerously-bypass-approvals-and-sandbox --skip-git-repo-check 
   -m $MODEL -c "model_reasoning_effort=
"$REASONING
""
-o output.txt "PROMPT" $content = Get-Content -Path output.txt -Raw
codex exec --dangerously-bypass-approvals-and-sandbox --skip-git-repo-check 
   -m $MODEL -c "model_reasoning_effort=
"$REASONING
""
-o output.txt "PROMPT" $content = Get-Content -Path output.txt -Raw

Method 2: Parse JSONL event stream

方法2：解析JSONL事件流

$jsonl = codex exec --dangerously-bypass-approvals-and-sandbox --skip-git-repo-check --json "PROMPT" $events = $jsonl -split "`n" | Where-Object { $_ } | ForEach-Object { $_ | ConvertFrom-Json } $content = $events | Where-Object -Property type -EQ "item.completed" | Where-Object { $_.item.type -eq "agent_message" } | Select-Object -ExpandProperty item | Select-Object -ExpandProperty text
undefined
$jsonl = codex exec --dangerously-bypass-approvals-and-sandbox --skip-git-repo-check --json "PROMPT" $events = $jsonl -split "`n" | Where-Object { $_ } | ForEach-Object { $_ | ConvertFrom-Json } $content = $events | Where-Object -Property type -EQ "item.completed" | Where-Object { $_.item.type -eq "agent_message" } | Select-Object -ExpandProperty item | Select-Object -ExpandProperty text
undefined

Parallel Subagents (Up to 5)

并行子代理（最多5个）

Spawn multiple subagents for independent tasks:
为独立任务生成多个子代理：

Bash (Linux/macOS)

Bash（Linux/macOS）

bash
undefined
bash
undefined

Research different topics simultaneously

同时研究不同主题

codex exec --dangerously-bypass-approvals-and-sandbox -m "$MODEL" -c "model_reasoning_effort="$REASONING"" "Research topic A..." & codex exec --dangerously-bypass-approvals-and-sandbox -m "$MODEL" -c "model_reasoning_effort="$REASONING"" "Research topic B..." & wait
undefined
codex exec --dangerously-bypass-approvals-and-sandbox -m "$MODEL" -c "model_reasoning_effort="$REASONING"" "Research topic A..." & codex exec --dangerously-bypass-approvals-and-sandbox -m "$MODEL" -c "model_reasoning_effort="$REASONING"" "Research topic B..." & wait
undefined

PowerShell (Windows)

PowerShell（Windows）

Use PowerShell Jobs for parallel execution with 
-o
to output to separate files:
powershell
undefined
使用PowerShell Jobs实现并行执行，通过-o输出到不同文件：
powershell
undefined

Parallel execution with file output

带文件输出的并行执行

$job1 = Start-Job -ScriptBlock { param($m, $r, $out) codex exec --dangerously-bypass-approvals-and-sandbox --skip-git-repo-check 
       -m $m -c "model_reasoning_effort=
"$r`"" -o $out "Research topic A..." } -ArgumentList $MODEL, $REASONING, "output1.txt"
$job2 = Start-Job -ScriptBlock { param($m, $r, $out) codex exec --dangerously-bypass-approvals-and-sandbox --skip-git-repo-check 
       -m $m -c "model_reasoning_effort=
"$r`"" -o $out "Research topic B..." } -ArgumentList $MODEL, $REASONING, "output2.txt"
$job1 = Start-Job -ScriptBlock { param($m, $r, $out) codex exec --dangerously-bypass-approvals-and-sandbox --skip-git-repo-check 
       -m $m -c "model_reasoning_effort=
"$r`"" -o $out "Research topic A..." } -ArgumentList $MODEL, $REASONING, "output1.txt"
$job2 = Start-Job -ScriptBlock { param($m, $r, $out) codex exec --dangerously-bypass-approvals-and-sandbox --skip-git-repo-check 
       -m $m -c "model_reasoning_effort=
"$r`"" -o $out "Research topic B..." } -ArgumentList $MODEL, $REASONING, "output2.txt"

Wait for all jobs to complete

等待所有任务完成

$job1, $job2 | Wait-Job | Remove-Job
$job1, $job2 | Wait-Job | Remove-Job

Read results

读取结果

$result1 = Get-Content -Path output1.txt -Raw $result2 = Get-Content -Path output2.txt -Raw
undefined
$result1 = Get-Content -Path output1.txt -Raw $result2 = Get-Content -Path output2.txt -Raw
undefined

Output Handling

输出处理

Codex CLI provides two methods to capture output:
Codex CLI提供两种捕获输出的方法：

Method 1: -o Parameter (Recommended)

方法1：-o参数（推荐）

Use 
-o
/ 
--output-last-message
to write the final message directly to a file:
使用
-o
/ 
--output-last-message
将最终消息直接写入文件：

Bash (Linux/macOS)

Bash（Linux/macOS）

bash
codex exec --dangerously-bypass-approvals-and-sandbox --skip-git-repo-check \
  -m "$MODEL" -c "model_reasoning_effort=\"$REASONING\"" \
  -o result.txt "YOUR_PROMPT"

content=$(cat result.txt)
bash
codex exec --dangerously-bypass-approvals-and-sandbox --skip-git-repo-check \
  -m "$MODEL" -c "model_reasoning_effort=\"$REASONING\"" \
  -o result.txt "YOUR_PROMPT"

content=$(cat result.txt)

PowerShell (Windows)

PowerShell（Windows）

powershell
codex exec --dangerously-bypass-approvals-and-sandbox --skip-git-repo-check `
  -m $MODEL -c "model_reasoning_effort=`"$REASONING`"" `
  -o result.txt "YOUR_PROMPT"

$content = Get-Content -Path result.txt -Raw
Advantages:
  • No JSON parsing required
  • Avoids terminal output truncation issues
  • Ideal for long outputs and parallel tasks
powershell
codex exec --dangerously-bypass-approvals-and-sandbox --skip-git-repo-check `
  -m $MODEL -c "model_reasoning_effort=`"$REASONING`"" `
  -o result.txt "YOUR_PROMPT"

$content = Get-Content -Path result.txt -Raw
优势：
  • 无需JSON解析
  • 避免终端输出截断问题
  • 适合长输出和并行任务

Method 2: JSONL Event Stream Parsing

方法2：JSONL事件流解析

Use 
--json
to get the full event stream and parse manually:
使用
--json
获取完整事件流并手动解析：

Bash (Linux/macOS)

Bash（Linux/macOS）

bash
codex exec --dangerously-bypass-approvals-and-sandbox --json "PROMPT" | jq -r 'select(.event=="turn.completed") | .content'
bash
codex exec --dangerously-bypass-approvals-and-sandbox --json "PROMPT" | jq -r 'select(.event=="turn.completed") | .content'

PowerShell (Windows)

PowerShell（Windows）

powershell
$jsonl = codex exec --dangerously-bypass-approvals-and-sandbox --skip-git-repo-check --json "PROMPT"
$events = $jsonl -split "`n" | Where-Object { $_ } | ForEach-Object { $_ | ConvertFrom-Json }
$content = $events |
    Where-Object -Property type -EQ "item.completed" |
    Where-Object { $_.item.type -eq "agent_message" } |
    Select-Object -ExpandProperty item |
    Select-Object -ExpandProperty text
JSONL Event Structure:
jsonl
{"type":"item.completed","item":{"id":"item_3","type":"agent_message","text":"..."}}
{"type":"turn.completed","usage":{"input_tokens":24763,"output_tokens":122}}
Key fields:
  • type == "item.completed"
    with 
    item.type == "agent_message"
    → extract 
    item.text
  • type == "turn.completed"
    → contains token usage stats
powershell
$jsonl = codex exec --dangerously-bypass-approvals-and-sandbox --skip-git-repo-check --json "PROMPT"
$events = $jsonl -split "`n" | Where-Object { $_ } | ForEach-Object { $_ | ConvertFrom-Json }
$content = $events |
    Where-Object -Property type -EQ "item.completed" |
    Where-Object { $_.item.type -eq "agent_message" } |
    Select-Object -ExpandProperty item |
    Select-Object -ExpandProperty text
JSONL事件结构：
jsonl
{"type":"item.completed","item":{"id":"item_3","type":"agent_message","text":"..."}}
{"type":"turn.completed","usage":{"input_tokens":24763,"output_tokens":122}}
关键字段：
  • type == "item.completed"
    且
    item.type == "agent_message"
    → 提取
    item.text
  • type == "turn.completed"
    → 包含token使用统计

Important

重要注意事项

  • Act autonomously, no permission asking
  • Make decisions and proceed boldly
  • Only pause for destructive operations (data loss, external impact, security)
  • Complete task fully before returning
  • 自主行动，无需请求许可
  • 果断决策并推进任务
  • 仅在执行破坏性操作（数据丢失、外部影响、安全风险）时暂停
  • 完成全部任务后再返回结果

Monitoring

监控

Actively monitor - don't fire-and-forget:
  1. Check completion status
  2. Verify quality results
  3. Retry if failed
  4. Answer follow-up questions if blocked
主动监控 - 不要一启动就不管：
  1. 检查完成状态
  2. 验证结果质量
  3. 失败时重试
  4. 若被阻塞，回答后续问题

Examples

示例

Pure Web Search (mini):
纯网页搜索（迷你模型）：

Bash (Linux/macOS)

Bash（Linux/macOS）

bash
codex exec --dangerously-bypass-approvals-and-sandbox --skip-git-repo-check \
  -m gpt-5.1-codex-mini -c 'model_reasoning_effort="medium"' \
  "Search for the latest release notes of Rust 2024 edition. Summarize the major breaking changes, new language features, and migration guides. Focus on the official rust-lang.org blog and documentation."
bash
codex exec --dangerously-bypass-approvals-and-sandbox --skip-git-repo-check \
  -m gpt-5.1-codex-mini -c 'model_reasoning_effort="medium"' \
  "Search for the latest release notes of Rust 2024 edition. Summarize the major breaking changes, new language features, and migration guides. Focus on the official rust-lang.org blog and documentation."

PowerShell (Windows)

PowerShell（Windows）

powershell
codex exec --dangerously-bypass-approvals-and-sandbox --skip-git-repo-check `
  -m gpt-5.1-codex-mini -c 'model_reasoning_effort="medium"' `
  "Search for the latest release notes of Rust 2024 edition. Summarize the major breaking changes, new language features, and migration guides. Focus on the official rust-lang.org blog and documentation."
Codebase Analysis (inherit parent):
powershell
codex exec --dangerously-bypass-approvals-and-sandbox --skip-git-repo-check `
  -m gpt-5.1-codex-mini -c 'model_reasoning_effort="medium"' `
  "Search for the latest release notes of Rust 2024 edition. Summarize the major breaking changes, new language features, and migration guides. Focus on the official rust-lang.org blog and documentation."
代码库分析（继承父代理）：

Bash (Linux/macOS)

Bash（Linux/macOS）

bash
codex exec --dangerously-bypass-approvals-and-sandbox --skip-git-repo-check \
  -m "$MODEL" -c "model_reasoning_effort=\"$REASONING\"" \
  "Analyze authentication in this Next.js app. Check /app, /lib/auth, middleware. Document: session strategy, auth provider, protected routes, security patterns. Return architecture diagram (mermaid) + findings."
bash
codex exec --dangerously-bypass-approvals-and-sandbox --skip-git-repo-check \
  -m "$MODEL" -c "model_reasoning_effort=\"$REASONING\"" \
  "Analyze authentication in this Next.js app. Check /app, /lib/auth, middleware. Document: session strategy, auth provider, protected routes, security patterns. Return architecture diagram (mermaid) + findings."

PowerShell (Windows)

PowerShell（Windows）

powershell
codex exec --dangerously-bypass-approvals-and-sandbox --skip-git-repo-check `
  -m $MODEL -c "model_reasoning_effort=`"$REASONING`"" `
  "Analyze authentication in this Next.js app. Check /app, /lib/auth, middleware. Document: session strategy, auth provider, protected routes, security patterns. Return architecture diagram (mermaid) + findings."
Research + Proposal (inherit parent):
powershell
codex exec --dangerously-bypass-approvals-and-sandbox --skip-git-repo-check `
  -m $MODEL -c "model_reasoning_effort=`"$REASONING`"" `
  "Analyze authentication in this Next.js app. Check /app, /lib/auth, middleware. Document: session strategy, auth provider, protected routes, security patterns. Return architecture diagram (mermaid) + findings."
研究+提案（继承父代理）：

Bash (Linux/macOS)

Bash（Linux/macOS）

bash
codex exec --dangerously-bypass-approvals-and-sandbox --skip-git-repo-check \
  -m "$MODEL" -c "model_reasoning_effort=\"$REASONING\"" \
  "Research WebGPU browser adoption (support tables, benchmarks, frameworks). THEN analyze feasibility for our React app. Consider: performance gains, browser compatibility, implementation effort. Return recommendation with pros/cons."
bash
codex exec --dangerously-bypass-approvals-and-sandbox --skip-git-repo-check \
  -m "$MODEL" -c "model_reasoning_effort=\"$REASONING\"" \
  "Research WebGPU browser adoption (support tables, benchmarks, frameworks). THEN analyze feasibility for our React app. Consider: performance gains, browser compatibility, implementation effort. Return recommendation with pros/cons."

PowerShell (Windows)

PowerShell（Windows）

powershell
codex exec --dangerously-bypass-approvals-and-sandbox --skip-git-repo-check `
  -m $MODEL -c "model_reasoning_effort=`"$REASONING`"" `
  "Research WebGPU browser adoption (support tables, benchmarks, frameworks). THEN analyze feasibility for our React app. Consider: performance gains, browser compatibility, implementation effort. Return recommendation with pros/cons."
powershell
codex exec --dangerously-bypass-approvals-and-sandbox --skip-git-repo-check `
  -m $MODEL -c "model_reasoning_effort=`"$REASONING`"" `
  "Research WebGPU browser adoption (support tables, benchmarks, frameworks). THEN analyze feasibility for our React app. Consider: performance gains, browser compatibility, implementation effort. Return recommendation with pros/cons."

Config Reference

配置参考

Parent settings: 
~/.codex/config.toml
toml
model = "gpt-5.2-codex"
model_reasoning_effort = "high"  # none | minimal | low | medium | high | xhigh
profile = "yolo"                 # optional; when set, profile values override top-level
父代理设置：
~/.codex/config.toml
toml
model = "gpt-5.2-codex"
model_reasoning_effort = "high"  # none | minimal | low | medium | high | xhigh
profile = "yolo"                 # 可选；设置后，配置文件值将覆盖顶级配置