solo
Use for a small feature, refactor, or UI tweak. You get project context, the narrowest meaningful validation, and one diff-only self-check without another reviewer.
diff --git a/README.md b/README.md index d1eb0aa..7b31c67 100644 --- a/README.md +++ b/README.md @@ -17,7 +17,7 @@
@@ -46,7 +46,7 @@ mancode installs three things: 1. **Hooks** that inject project context, design tokens, and YAGNI checks into agent prompts. -2. **Skills / modes** for `solo`, `/mamba`, `/man`, `/manteam`, `/manps`, and +2. **Skills / modes** for `solo`, `/manba`, `/man`, `/manteam`, `/manps`, and `/mansolo`. 3. **Coaching-staff subagents**: Scout, Plan Coach, Head Coach, Film Analyst (Offense), and Film Analyst (Defense). @@ -80,15 +80,16 @@ After initialization, keep using your coding agent normally. `solo` mode runs by default: practice day, no ceremony. Use `/man` when a task needs planning, testing, and multi-agent review: playoffs, every possession counts. -Invocation is surface-specific. Claude Code and Cursor use `/man`, `/mamba`, +Invocation is surface-specific. Claude Code and Cursor use `/man`, `/manba`, and the other slash-style mode names. Codex in the ChatGPT desktop app, CLI, or -IDE extension loads repo skills from `.agents/skills/`; `$man`, `$mamba`, and +IDE extension loads repo skills from `.agents/skills/`; `$man`, `$manba`, and the other `$` mentions are the portable explicit syntax. In the ChatGPT desktop app, enabled skills also appear in the slash-command list, so a discovered `man` skill can be selected there as `/man`. In CLI/IDE, use `$man` or `/skills`. These are agent skills, not deprecated custom prompts. See the official [skills](https://learn.chatgpt.com/docs/build-skills) and [slash-command](https://learn.chatgpt.com/docs/reference/slash-commands) docs. +Existing workflow metadata remains compatible and does not need migration. ## What Gets Installed @@ -201,7 +202,7 @@ slash commands in your AI coding agent's conversation: | Mode | Best For | What It Does | |---|---|---| | `solo` | Daily coding · practice day | Lightweight hooks, style awareness, YAGNI checks, and one bounded diff self-check | -| `/mamba` | Diagnosis and real validation · Mamba mentality | Reproduces defects, finds root causes, drives real user flows, and runs regression checks | +| `/manba` | Diagnosis and real validation · Mamba mentality | Reproduces defects, finds root causes, drives real user flows, and runs regression checks | | `/man` | Production or high-risk changes · playoffs | Full 9-step workflow with targeted or full risk-based review | | `/manteam` | Team projects · five on the floor, one mind | Shared memory, decisions, coordination, and Conventional Commits | | `/manps` | Cleanup and maintenance · preseason | Project health scan with Markdown and JSON reports | @@ -292,7 +293,7 @@ it should behave, and why previous decisions were made. ## Installation -**Status**: stable v0.3.5. Claude Code, Cursor, Codex in the ChatGPT desktop app +**Status**: stable v0.3.6. Claude Code, Cursor, Codex in the ChatGPT desktop app and CLI, and GitHub Copilot are supported. ZCode adapter support is included, with project skill discovery kept behind a verification gate before release. @@ -342,14 +343,14 @@ mancode install --minimal # Install only solo-mode essentials ```bash # Claude Code / Cursor -/mamba # Diagnose bugs and validate real user flows +/manba # Diagnose bugs and validate real user flows /man # Full 9-step workflow with bounded risk-based review /manps # Project health check /manteam # Team mode and shared memory /mansolo # Return to solo mode # Codex in ChatGPT desktop / CLI / IDE -$mamba +$manba $man $manps $manteam @@ -364,7 +365,7 @@ mancode status mancode status --json mancode install
mancode
- Documentation / v0.3.5
+ Documentation / v0.3.6
Practice to playoffs
mancode is a local-first workflow harness for the coding agent you already use. It adds project context, task intensity, durable artifacts, and review discipline without replacing Claude Code, Cursor, Codex, Copilot, or ZCode.
-Use solo for day-to-day work. It performs one bounded diff self-check with no extra reviewer loop. Escalate to /mamba for diagnosis and /man when a production change deserves a plan, tests, and risk-based review.
Use solo for day-to-day work. It performs one bounded diff self-check with no extra reviewer loop. Escalate to /manba for diagnosis and /man when a production change deserves a plan, tests, and risk-based review.
mancode is deliberately not a mandatory ceremony layer. Choose the smallest mode that creates enough confidence for the task.
Use for a small feature, refactor, or UI tweak. You get project context, the narrowest meaningful validation, and one diff-only self-check without another reviewer.
Use when the right answer depends on reproduction and real validation—not a plausible patch.
Use when the right answer depends on reproduction and real validation—not a plausible patch.
Use when planning, verification, and independent review need to be durable and visible.
Use shared memory for a team; use health scans before a cleanup or dependency-maintenance pass.
The five working modes cover different kinds of engineering pressure. /mansolo is the reset command that returns the project to the default solo mode.
| Mode | Best for | What changes |
|---|---|---|
solo | Daily coding | Lightweight hooks, style awareness, YAGNI checks, and one bounded diff self-check. |
/mamba | Diagnosis | Reproduce the failure, find the cause, validate real user paths, and run regression checks. |
/manba | Diagnosis | Reproduce the failure, find the cause, validate real user paths, and run regression checks. |
/man | High-stakes work | Full nine-step workflow with a plan gate, targeted/full risk review, and one remediation round. |
/manteam | Team projects | Shared memory, decision records, coordination, and Conventional Commits guidance. |
/manps | Maintenance | Deterministic project-health scan with Markdown and JSON reports. |
Claude Code and Cursor expose slash-style commands. Codex in the ChatGPT desktop app, CLI, and IDE reads repository skills, so $man, $mamba, and related mentions are the portable explicit syntax.
Claude Code and Cursor expose slash-style commands. Codex in the ChatGPT desktop app, CLI, and IDE reads repository skills, so $man, $manba, and related mentions are the portable explicit syntax.
# Claude Code / Cursor
-/mamba
+/manba
/man
/manteam
# Codex app / CLI / IDE
-$mamba
+$manba
$man
-$manteamShows the current mode, detected stack, installed platforms, and readiness. Claude Code installations also report hook registration and estimated injection size.
--json for scripts or CI.Installs or repairs one platform adapter after initialization.
claude-code, cursor, codex, copilot, zcode.--force to rewrite generated configuration.--minimal for solo-mode essentials only.Lists the adapters this version knows how to install. Use this before scripting platform selection.
Creates and manages validated metadata for man, mamba, and manteam workflow runs.
create <mode> "<task>" starts a record.update <taskId> advances steps or status.review <taskId> … enforces required domains, blockers, and one remediation round.list, show, and clean inspect and maintain records.Creates and manages validated metadata for man, manba, and manteam workflow runs.
create <mode> "<task>" starts a record.update <taskId> advances steps or status.review <taskId> … enforces required domains, blockers, and one remediation round.list, show, and clean inspect and maintain records.Runs a deterministic preseason health scan. Areas include deps, security, dead-code, and config.
.mancode/.Refreshes the project profile and, when UI assets exist, rescans design tokens.
project-profile.json and style-tokens.json.mancode install <platform> --force for static-instruction adapters.Prints the installed CLI version. Use it when reporting an issue or validating an update.
A linked mamba child may only be created while its parent workflow is active at Step 6. New governed workflows also cannot complete until their required review domains are recorded and all blocker IDs are resolved.
A linked manba child may only be created while its parent workflow is active at Step 6. New governed workflows also cannot complete until their required review domains are recorded and all blocker IDs are resolved.
mancode
- 文档中心 / v0.3.5
+ 文档中心 / v0.3.6
训练到季后赛
mancode 是加在既有编码 Agent 之上的本地优先工作流层。它不替代 Claude Code、Cursor、Codex、Copilot 或 ZCode,而是增加项目上下文、任务强度、可追溯产物和审查纪律。
-日常任务用 solo:只做一次 diff 自检,不启动额外 reviewer 循环;需要定位根因时切换 /mamba;生产级或高风险改动再使用 /man,获得计划、验证和有界风险审查。
日常任务用 solo:只做一次 diff 自检,不启动额外 reviewer 循环;需要定位根因时切换 /manba;生产级或高风险改动再使用 /man,获得计划、验证和有界风险审查。
如果你知道当前项目要面向哪个 Agen
mancode 的目的不是让所有任务都变重,而是为不同风险选择足够的工程纪律。 适合小功能、重构或 UI 微调。使用最窄验证和一次 diff 自检,不调用额外 reviewer。 当正确答案依赖复现与真实验证,而不是一个看起来合理的补丁时使用。 当计划、测试和有界风险审查需要被记录、复查和交接时使用。 团队共享记忆;维护前先执行健康扫描,避免拍脑袋式清理。 适合小功能、重构或 UI 微调。使用最窄验证和一次 diff 自检,不调用额外 reviewer。 当正确答案依赖复现与真实验证,而不是一个看起来合理的补丁时使用。 当计划、测试和有界风险审查需要被记录、复查和交接时使用。 团队共享记忆;维护前先执行健康扫描,避免拍脑袋式清理。 不要因为任务很长就自动使用 五个工作模式对应不同工程压力。 Claude Code 与 Cursor 使用 slash 风格命令;ChatGPT 桌面端 Codex、Codex CLI 和 IDE 从项目 skill 中读取,因此 Claude Code 与 Cursor 使用 slash 风格命令;ChatGPT 桌面端 Codex、Codex CLI 和 IDE 从项目 skill 中读取,因此 推荐用法
solo
/mamba
/man
/manteam · /manps
solo
/manba
/man
/manteam · /manps
/man。当错误代价高、难以回滚,或事后难以审查时,再升级到季后赛模式。模式
/mansolo 是回到默认 solo 模式的重置命令。
-
- 模式 适合场景 会做什么 solo日常编码 轻量 hooks、YAGNI、最窄验证和一次受限 diff 自检。 /mamba诊断 复现缺陷、定位根因、驱动真实用户路径并回归验证。 /man高风险改动 完整九步工作流、计划关卡、定向/完整风险审查和一轮修复。 /manteam团队项目 共享记忆、决策记录、协作和 Conventional Commits 指导。
+ /manps维护 输出 Markdown 与 JSON 的确定性项目健康报告。 solo日常编码 轻量 hooks、YAGNI、最窄验证和一次受限 diff 自检。 /manba诊断 复现缺陷、定位根因、驱动真实用户路径并回归验证。 /man高风险改动 完整九步工作流、计划关卡、定向/完整风险审查和一轮修复。 /manteam团队项目 共享记忆、决策记录、协作和 Conventional Commits 指导。 /manps维护 输出 Markdown 与 JSON 的确定性项目健康报告。 怎样调用
$man、$mamba 等是稳定的显式调用方式。怎样调用
$man、$manba 等是稳定的显式调用方式。# Claude Code / Cursor
-/mamba
+/manba
/man
/manteam
# Codex app / CLI / IDE
-$mamba
+$manba
$man
-$manteam
以下命令除非另有说明,都应该在已初始化
显示当前模式、检测到的技术栈、已安装平台和就绪状态。Claude Code 安装还会显示 hook 注册与预计注入大小。 初始化后安装或修复单个平台适配器。 列出当前版本可以安装的适配器。自动化选择平台前可以先运行它。 为 为 运行确定性的季前赛健康检查。可指定 刷新项目 profile;检测到 UI 时重新扫描设计 token。 打印当前安装的 CLI 版本。提交 issue 或确认升级时使用。mancode status [--json]
检查--json 适合脚本或 CI。mancode install <platform>
适配器claude-code、cursor、codex、copilot、zcode。--force 重写生成配置。--minimal 只安装 solo 必需项。mancode list-platforms
发现mancode workflow …
记录man、mamba 与 manteam 创建和维护受校验的工作流元数据。create <mode> "<task>" 创建记录。update <taskId> 更新步骤或状态。review <taskId> … 约束审查领域、blocker 和唯一一轮修复。list、show、clean 用于检查和维护。mancode workflow …
记录man、manba 与 manteam 创建和维护受校验的工作流元数据。create <mode> "<task>" 创建记录。update <taskId> 更新步骤或状态。review <taskId> … 约束审查领域、blocker 和唯一一轮修复。list、show、clean 用于检查和维护。mancode manps [area]
扫描deps、security、dead-code、config。.mancode/。mancode refresh-style
刷新project-profile.json 与 style-tokens.json。mancode install <platform> --force。mancode version
版本
以下命令除非另有说明,都应该在已初始化 $ mancode workflow review <taskId> init --review-depth full $ mancode workflow review <taskId> complete --review-domain quality --report film-report-1.md --blockers Q1 $ mancode workflow review <taskId> remediate --resolved Q1 -$ mancode workflow create mamba "verify auth regression" --parent-task <taskId> +$ mancode workflow create manba "verify auth regression" --parent-task <taskId> $ mancode workflow show <taskId> --json -$ mancode workflow clean --older-than 30d --dry-run
关联的 mamba 子任务只能在父工作流处于 Step 6 且仍在进行时创建。新的严格 workflow 也必须完成所需审查领域并清零 blocker 后才能完成。
关联的 manba 子任务只能在父工作流处于 Step 6 且仍在进行时创建。新的严格 workflow 也必须完成所需审查领域并清零 blocker 后才能完成。
mancode 不假装所有工具都有相同扩展点。Claude Code 获得最完整的原生能力;其他平台通过持久 rules、skills 或受控 instructions 适配。
| 平台 | 安装内容 | 说明 |
|---|---|---|
| Claude Code | Hooks、skills、subagents、workflow | 完整集成:会话启动与提交提示时可注入项目上下文。 |
| Cursor | .cursor/rules/*.mdc | 通过 rules 保存项目指导与模式描述。 |
| Codex | AGENTS.md + .agents/skills/ | 在 ChatGPT 桌面端、CLI 与 IDE 使用 $man* 触发 skill。 |
| GitHub Copilot | .github/copilot-instructions.md | 受控 instruction block;当前没有等价 hook 注入。 |
| ZCode | AGENTS.md + 预备项目 skills | 项目 skill 发现与 slash command 路径仍是发布验证门槛。 |