Back to Details

post-mortem

Compare original and translation side by side

🇺🇸

Original

English
🇨🇳

Translation

Chinese

Post-Mortem Skill

Post-Mortem Skill

Purpose: Wrap up completed work — validate it shipped correctly and extract learnings.
Two steps:
  1. /council validate
    — Did we implement it correctly?
  2. /retro
    — What did we learn?
用途： 收尾已完成的工作——验证成果是否正确交付，并提取经验教训。
两个步骤：
  1. /council validate
    —— 我们的实施是否符合要求？
  2. /retro
    —— 我们学到了什么？

Quick Start

快速开始

bash
/post-mortem                    # wraps up recent work
/post-mortem epic-123           # wraps up specific epic
/post-mortem --quick recent     # fast inline wrap-up, no spawning
/post-mortem --deep recent      # thorough council review
/post-mortem --mixed epic-123   # cross-vendor (Claude + Codex)
/post-mortem --explorers=2 epic-123  # deep investigation before judging
/post-mortem --debate epic-123      # two-round adversarial review
bash
/post-mortem                    # 收尾近期工作
/post-mortem epic-123           # 收尾特定epic
/post-mortem --quick recent     # 快速内联收尾，不启动独立进程
/post-mortem --deep recent      # 全面的Council评审
/post-mortem --mixed epic-123   # 跨供应商（Claude + Codex）评审
/post-mortem --explorers=2 epic-123  # 评审前先进行深度调查
/post-mortem --debate epic-123      # 两轮对抗式评审

Execution Steps

执行步骤

Step 1: Identify Completed Work

步骤1：确定已完成的工作

If epic/issue ID provided: Use it directly.
If no ID: Find recently completed work:
bash
undefined
如果提供了epic/问题ID： 直接使用该ID。
如果未提供ID： 查找近期已完成的工作：
bash
undefined

Check for closed beads

查看已关闭的bead

bd list --status closed --since "7 days ago" 2>/dev/null | head -5
bd list --status closed --since "7 days ago" 2>/dev/null | head -5

Or check recent git activity

或查看近期git活动

git log --oneline --since="7 days ago" | head -10
undefined
git log --oneline --since="7 days ago" | head -10
undefined

Step 2: Load the Original Plan/Spec

步骤2：加载原始计划/规格说明

Before invoking council, load the original plan for comparison:
  1. If epic/issue ID provided: 
    bd show <id>
    to get the spec/description
  2. Search for plan doc: 
    ls .agents/plans/ | grep <target-keyword>
  3. Check git log: 
    git log --oneline | head -10
    to find the relevant bead reference
If a plan is found, include it in the council packet's 
context.spec
field:
json
{
  "spec": {
    "source": "bead na-0042",
    "content": "<the original plan/spec text>"
  }
}
调用Council之前，加载原始计划以便进行对比：
  1. 如果提供了epic/问题ID： 使用
    bd show <id>
    获取规格说明/描述
  2. 搜索计划文档： 
    ls .agents/plans/ | grep <目标关键词>
  3. 查看git日志： 
    git log --oneline | head -10
    查找相关的bead引用
如果找到计划，将其包含在Council数据包的
context.spec
字段中：
json
{
  "spec": {
    "source": "bead na-0042",
    "content": "<原始计划/规格说明文本>"
  }
}

Step 3: Council Validates the Work

步骤3：Council验证工作成果

Run 
/council
with the retrospective preset and always 3 judges:
/council --deep --preset=retrospective validate <epic-or-recent>
Default (3 judges with retrospective perspectives):
  • plan-compliance
    : What was planned vs what was delivered? What's missing? What was added?
  • tech-debt
    : What shortcuts were taken? What will bite us later? What needs cleanup?
  • learnings
    : What patterns emerged? What should be extracted as reusable knowledge?
Post-mortem always uses 3 judges (
--deep
) because completed work deserves thorough review.
The plan/spec content is injected into the council packet context so the 
plan-compliance
judge can compare planned vs delivered.
With --quick (inline, no spawning):
/council --quick validate <epic-or-recent>
Single-agent structured review. Fast wrap-up without spawning.
With debate mode:
/post-mortem --debate epic-123
Enables adversarial two-round review for post-implementation validation. Use for high-stakes shipped work where missed findings have production consequences. See 
/council
docs for full --debate details.
Advanced options (passed through to council):
  • --mixed
    — Cross-vendor (Claude + Codex) with retrospective perspectives
  • --preset=<name>
    — Override with different personas (e.g., 
    --preset=ops
    for production readiness)
  • --explorers=N
    — Each judge spawns N explorers to investigate the implementation deeply before judging
  • --debate
    — Two-round adversarial review (judges critique each other's findings before final verdict)
使用回顾预设并指定3名评审者运行
/council
：
/council --deep --preset=retrospective validate <epic或近期工作>
默认配置（3名带回顾视角的评审者）：
  • plan-compliance
    ：计划内容与交付成果有何差异？缺少了什么？新增了什么？
  • tech-debt
    ：采取了哪些捷径？哪些问题会在后续造成影响？哪些内容需要清理？
  • learnings
    ：出现了哪些模式？哪些内容可以提炼为可复用的知识？
Post-mortem始终使用3名评审者（
--deep
），因为已完成的工作值得全面评审。
计划/规格说明内容会被注入到Council数据包的上下文中，以便
plan-compliance
评审者对比计划与交付成果。
使用--quick（内联模式，不启动独立进程）：
/council --quick validate <epic或近期工作>
单Agent结构化评审，快速收尾无需启动独立进程。
使用辩论模式：
/post-mortem --debate epic-123
为实施后验证启用两轮对抗式评审。适用于高风险已交付工作，若遗漏问题会对生产环境造成影响。查看
/council
文档了解
--debate
的完整细节。
高级选项（传递给Council）：
  • --mixed
    —— 跨供应商（Claude + Codex）并采用回顾视角
  • --preset=<名称>
    —— 用不同角色覆盖默认配置（例如
    --preset=ops
    用于生产就绪性评审）
  • --explorers=N
    —— 每位评审者在评审前启动N个探索者深入调查实施细节
  • --debate
    —— 两轮对抗式评审（评审者在给出最终结论前互相批判对方的发现）

Step 4: Extract Learnings

步骤4：提取经验教训

Run 
/retro
to capture what we learned:
/retro <epic-or-recent>
Retro captures:
  • What went well?
  • What was harder than expected?
  • What would we do differently?
  • Patterns to reuse?
  • Anti-patterns to avoid?
Error Handling:
FailureBehavior
Council failsStop, report council error, no retro
Retro failsProceed, report learnings as "⚠️ SKIPPED: retro unavailable"
Both succeedFull post-mortem with council + learnings
Post-mortem always completes if council succeeds. Retro is optional enrichment.
运行
/retro
记录我们学到的内容：
/retro <epic或近期工作>
Retro记录内容：
  • 哪些方面做得好？
  • 哪些方面比预期更困难？
  • 下次我们会怎么做？
  • 可复用的模式？
  • 需要避免的反模式？
错误处理：
失败场景行为
Council执行失败停止操作，报告Council错误，不执行Retro
Retro执行失败继续操作，记录经验教训为"⚠️ 已跳过：Retro不可用"
两者均执行成功完整的Post-Mortem流程，包含Council验证和经验教训提取
只要Council执行成功，Post-Mortem流程就会完成。Retro是可选的增强步骤。

Step 5: Write Post-Mortem Report

步骤5：编写Post-Mortem报告

Write to: 
.agents/council/YYYY-MM-DD-post-mortem-<topic>.md
markdown
undefined
保存路径： 
.agents/council/YYYY-MM-DD-post-mortem-<主题>.md
markdown
undefined

Post-Mortem: <Epic/Topic>

Post-Mortem：<Epic/主题>

Date: YYYY-MM-DD Epic: <epic-id or "recent"> Duration: <how long it took>
日期： YYYY-MM-DD Epic： <epic-id或"近期工作"> 时长： <耗时多久>

Council Verdict: PASS / WARN / FAIL

Council评审结论：通过 / 警告 / 失败

JudgeVerdictKey Finding
Plan-Compliance......
Tech-Debt......
Learnings......
评审者结论关键发现
计划合规性......
技术债务......
经验教训......

Implementation Assessment

实施评估

<council summary>
<Council总结>

Concerns

问题与担忧

<any issues found>
<发现的所有问题>

Learnings (from /retro)

经验教训（来自/retro）

What Went Well

做得好的方面

  • ...
  • ...

What Was Hard

困难之处

  • ...
  • ...

Do Differently Next Time

改进措施

  • ...
  • ...

Patterns to Reuse

可复用模式

  • ...
  • ...

Anti-Patterns to Avoid

需避免的反模式

  • ...
  • ...

Status

状态

[ ] CLOSED - Work complete, learnings captured [ ] FOLLOW-UP - Issues need addressing (create new beads)
undefined
[ ] 已关闭 - 工作完成，经验教训已记录 [ ] 待跟进 - 存在需要处理的问题（创建新的bead）
undefined

Step 6: Feed the Knowledge Flywheel

步骤6：为知识飞轮提供数据

Post-mortem automatically feeds learnings into the flywheel:
bash
mkdir -p .agents/knowledge/pending

if command -v ao &>/dev/null; then
  ao forge index .agents/learnings/ 2>/dev/null
  echo "Learnings indexed in knowledge flywheel"
else
  # Retro already wrote to .agents/learnings/ — copy to pending for future import
  cp .agents/learnings/YYYY-MM-DD-*.md .agents/knowledge/pending/ 2>/dev/null
  echo "Note: Learnings saved to .agents/knowledge/pending/ (install ao for auto-indexing)"
fi
Post-Mortem会自动将经验教训输入到知识飞轮：
bash
mkdir -p .agents/knowledge/pending

if command -v ao &>/dev/null; then
  ao forge index .agents/learnings/ 2>/dev/null
  echo "经验教训已索引到知识飞轮"
else
  # Retro已将内容写入.agents/learnings/ —— 复制到pending目录待后续导入
  cp .agents/learnings/YYYY-MM-DD-*.md .agents/knowledge/pending/ 2>/dev/null
  echo "注意：经验教训已保存到.agents/knowledge/pending/（安装ao可实现自动索引）"
fi

Step 7: Report to User

步骤7：向用户反馈

Tell the user:
  1. Council verdict on implementation
  2. Key learnings
  3. Any follow-up items
  4. Location of post-mortem report
  5. Knowledge flywheel status
告知用户以下信息：
  1. Council对实施成果的评审结论
  2. 关键经验教训
  3. 任何待跟进事项
  4. Post-Mortem报告的位置
  5. 知识飞轮的状态

Integration with Workflow

与工作流的集成

/plan epic-123
    │
    ▼
/pre-mortem (council on plan)
    │
    ▼
/implement
    │
    ▼
/vibe (council on code)
    │
    ▼
Ship it
    │
    ▼
/post-mortem              ← You are here
    │
    ├── Council validates implementation
    └── Retro extracts learnings
/plan epic-123
    │
    ▼
/pre-mortem（对计划进行Council评审）
    │
    ▼
/implement
    │
    ▼
/vibe（对代码进行Council评审）
    │
    ▼
交付上线
    │
    ▼
/post-mortem              ← 当前步骤
    │
    ├── Council验证实施成果
    └── Retro提取经验教训

Examples

示例

Wrap Up Recent Work

收尾近期工作

bash
/post-mortem
Validates recent commits, extracts learnings.
bash
/post-mortem
验证近期提交的成果，提取经验教训。

Wrap Up Specific Epic

收尾特定Epic

bash
/post-mortem epic-123
Council reviews epic-123 implementation, retro captures learnings.
bash
/post-mortem epic-123
Council评审epic-123的实施成果，Retro记录经验教训。

Thorough Review

全面评审

bash
/post-mortem --deep epic-123
3 judges review the epic.
bash
/post-mortem --deep epic-123
3名评审者对该epic进行评审。

Cross-Vendor Review

跨供应商评审

bash
/post-mortem --mixed epic-123
3 Claude + 3 Codex agents review the epic.
bash
/post-mortem --mixed epic-123
3名Claude + 3名Codex Agent对该epic进行评审。

Relationship to Other Skills

与其他Skill的关系

SkillWhenPurpose
/pre-mortem
Before implementationCouncil validates plan
/vibe
After codingCouncil validates code
/post-mortem
After shippingCouncil validates + extract learnings
/retro
AnytimeExtract learnings only
Skill使用时机用途
/pre-mortem
实施前Council验证计划
/vibe
编码后Council验证代码
/post-mortem
交付后Council验证成果 + 提取经验教训
/retro
任何时间仅提取经验教训

See Also

相关文档

  • skills/council/SKILL.md
    — Multi-model validation council
  • skills/retro/SKILL.md
    — Extract learnings
  • skills/vibe/SKILL.md
    — Council validates code
  • skills/pre-mortem/SKILL.md
    — Council validates plans
  • skills/council/SKILL.md
    —— 多模型验证Council
  • skills/retro/SKILL.md
    —— 经验教训提取
  • skills/vibe/SKILL.md
    —— Council代码验证
  • skills/pre-mortem/SKILL.md
    —— Council计划验证