Name: .codex
Author: WenJunDuan

.codex

1. **先搜后写**: augment-context-engine 搜现有实现 → 不可用时 grep -r 2. **先规后码**: Path B+ 任务必须先 plan.md → cunzhi 确认后才能写代码 3. **不确定就问**: 歧义/架构决策 → cunzhi 向用户确认, 不猜 4. **不破坏已有**: 改代码前读测试, 改后跑测试, 红了就修 5.

by WenJunDuan(0) 0 installs

Imported from WenJunDuan/Rlues (vibeCoding/old/03-config-agent-v8-v9/v9.0.8/.codex/AGENTS.md). Install upstream with npx skills add WenJunDuan/Rlues --skill .codex. Copyright stays with the author.

VibeCoding Kernel v9.0.8 — Codex CLI

铁律 (违反即失败)

先搜后写: augment-context-engine 搜现有实现 → 不可用时 grep -r
先规后码: Path B+ 任务必须先 plan.md → cunzhi 确认后才能写代码
不确定就问: 歧义/架构决策 → cunzhi 向用户确认, 不猜
不破坏已有: 改代码前读测试, 改后跑测试, 红了就修
只改需要改的: 不重构范围外代码, 不加未要求的功能
避免过度工程: 最简方案, YAGNI 优先
commit 粒度: 每个逻辑变更独立 commit, conventional commits

工作流

评估复杂度 → P.A.C.E. 路由 (读 .codex/workflows/pace.md)
按路径执行 RIPER-7 阶段 (读 .codex/workflows/riper-7.md)
按需加载 skills (读 .codex/skills/{name}/SKILL.md)

状态持久化

.ai_state/ 目录: session → design → plan → doing → verified → review → archive 模板位于 .codex/templates/ai-state/

MCP 降级

augment-context 不可用→grep | cunzhi 不可用→对话确认 | deepwiki 不可用→/search

Codex 专属能力

/review: 原生代码审查 (T 阶段集成)
/plan: 原生规划模式 (P 阶段集成)
multi_agent + collaboration_modes: 多代理并行 (Path C+, config.toml [features])
[agents] 角色定义: builder / reviewer / explorer (config.toml [agents.{role}])
spawn_agents_on_csv: 批量多代理分派
voice (spacebar): 语音输入 (实验性)
/model: gpt-5.3-codex (默认) / gpt-5.3-codex-spark (快速)

并行工作规范 (Path C/D)

读 plan.md → 识别无依赖任务组
为每组指派 agent 角色 (Codex 自动编排)
无依赖任务并行, 有依赖串行
同文件不并行修改
完成后合并 + /review 审查

# after Install: the listing is in your workspace's routing pool - a plan picks it for its slot curl -s -X POST https://api.opensmartroute.ai/api/v1/route -H 'Authorization: Bearer $OSR_API_KEY' -H 'Content-Type: application/json' -d '{"text": "...", "plan": true}'

Install .codex and it becomes one more candidate the router can pick - when it fits.

A listing is a routing target with a manifest: what it does, which domains it covers, what it costs and who publishes it. Once installed it sits beside your own models and tools, is scored like any other candidate for each request, and shows up in the trace when it wins. Ratings come from workspaces that installed it, one per account.

{ "ocm": "1", "id": "wenjunduan-rlues-codex", "kind": "skill", "name": ".codex", "description": "1. **先搜后写**: augment-context-engine 搜现有实现 → 不可用时 grep -r 2. **先规后码**: Path B+ 任务必须先 plan.md → cunzhi 确认后才能写代码 3. **不确定就问**: 歧义/架构决策 → cunzhi 向用户确认, 不猜 4. **不破坏已有**: 改代码前读测试, 改后跑测试, 红了就修 5. **只改需要改的**: 不重构范围外代码, 不加未要求的功能 6. **避免过度工程**: 最简方案, YAGNI 优先 7. **commit 粒度**: 每个逻辑变更独立 commit, conventional commits", "publisher": "WenJunDuan", "version": "1.0.0", "capabilities": { "domains": [ "general" ], "tags": [ "skill-md", "github-agents" ], "languages": [ "en" ] }, "quality_prior": 0.6, "examples": [ "1. **先搜后写**: augment-context-engine 搜现有实现 → 不可用时 grep -r 2. **先规后码**: Path B+ 任务必须先 plan.md → cunzhi 确认后才能写代码 3. **不确定就问**: 歧义/架构决策 → cunzhi 向用户确认, 不猜 4. **不破坏已有**: 改代码前读测试, 改后跑测试, 红了就修 5. **只改需要改的**: 不重构范围外代码, 不加未要求的功能 6. **避免过度工程**: 最简方案, YAGNI 优先 7. **commit 粒度**: 每个逻辑变更独立 commit, conventional commits" ], "primary": false, "metadata": { "source": { "provider": "github-agents", "repository": "https://github.com/WenJunDuan/Rlues", "path": "vibeCoding/old/03-config-agent-v8-v9/v9.0.8/.codex/AGENTS.md", "ref": "87b3920d0b8880a835fac4a25493391d5916d786", "url": "https://github.com/WenJunDuan/Rlues/blob/87b3920d0b8880a835fac4a25493391d5916d786/vibeCoding/old/03-config-agent-v8-v9/v9.0.8/.codex/AGENTS.md", "key": "WenJunDuan/Rlues/vibeCoding/old/03-config-agent-v8-v9/v9.0.8/.codex/AGENTS.md" } }, "instructions": "# VibeCoding Kernel v9.0.8 — Codex CLI\n\n## 铁律 (违反即失败)\n\n1. **先搜后写**: augment-context-engine 搜现有实现 → 不可用时 grep -r\n2. **先规后码**: Path B+ 任务必须先 plan.md → cunzhi 确认后才能写代码\n3. **不确定就问**: 歧义/架构决策 → cunzhi 向用户确认, 不猜\n4. **不破坏已有**: 改代码前读测试, 改后跑测试, 红了就修\n5. **只改需要改的**: 不重构范围外代码, 不加未要求的功能\n6. **避免过度工程**: 最简方案, YAGNI 优先\n7. **commit 粒度**: 每个逻辑变更独立 commit, conventional commits\n\n## 工作流\n\n1. 评估复杂度 → P.A.C.E. 路由 (读 .codex/workflows/pace.md)\n2. 按路径执行 RIPER-7 阶段 (读 .codex/workflows/riper-7.md)\n3. 按需加载 skills (读 .codex/skills/{name}/SKILL.md)\n\n## 状态持久化\n\n.ai_state/ 目录: session → design → plan → doing → verified → review", "cost": { "context_tokens": 297 } }

.codex

About

VibeCoding Kernel v9.0.8 — Codex CLI

铁律 (违反即失败)

工作流

状态持久化

MCP 降级

Codex 专属能力

并行工作规范 (Path C/D)

Use it

Install .codex and it becomes one more candidate the router can pick - when it fits.

Manifest

Reviews