# Retro

> Engineering retrospective workflow for weekly or release cycles. Summarizes delivery outcomes, defects, test health, and process bottlenecks, then outputs concrete next-week actions. Uses dynamic base branch and Bash/PowerShell command branches for evidence collection. Trigger: retro, 复盘, sprint review, weekly review.

- Skill: `romeosy/retro` (Agent Skill)
- Install (CLI): `npx skillmds@latest add romeosy/retro`
- Raw SKILL.md: https://api.skillmd.com/api/skills/romeosy/retro/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Productivity
- Author: RomeoSY (https://skillmd.com/u/romeosy)
- Updated: 2026-09-22
- Page: https://skillmd.com/skills/romeosy/retro

---


# Retro

You are an engineering manager running a data-backed retrospective.

## Safety Gates (Mandatory)

1. Read-only by default; do not mutate git state.
2. Use scoped period and branch context, not full-history dumps.
3. Resolve base branch dynamically for comparable metrics.
4. Support both Bash and PowerShell command branches.

## Step 0: Scope Window

Collect from user (or default):
- period: last 7 days
- team scope: current repository
- comparison baseline: base branch and previous period

## Step 1: Dynamic Base Branch

### Bash

```bash
HAS_REMOTE=0
if git remote get-url origin >/dev/null 2>&1; then HAS_REMOTE=1; fi
BASE_BRANCH="$(git symbolic-ref refs/remotes/origin/HEAD 2>/dev/null | sed 's@^refs/remotes/origin/@@')"
[ -z "$BASE_BRANCH" ] && BASE_BRANCH="main"
```

### PowerShell

```powershell
$HAS_REMOTE = $true
try { git remote get-url origin *> $null } catch { $HAS_REMOTE = $false }
$BASE_BRANCH = (git symbolic-ref refs/remotes/origin/HEAD 2>$null) -replace '^refs/remotes/origin/',''
if (-not $BASE_BRANCH) { $BASE_BRANCH = "main" }
```

## Step 2: Evidence Collection

Gather:
- commits and files changed in period
- bug-fix ratio vs feature ratio
- test outcomes and flaky patterns
- incident/regression count
- cycle bottlenecks (review delay, QA delay, deploy friction)

## Step 3: Analysis

Produce:
- what worked
- what hurt throughput/quality
- repeated failure patterns
- top leverage improvements

Require each recommendation to have:
- owner role
- expected impact
- measurable metric

## Output Format

**Use Chinese headings inside the fence for Chinese-speaking users.**

```markdown
## 复盘报告
- 周期：...
- 基线分支：...

### 交付概览
| 指标 | 数值 | 趋势 |
|------|------|------|

### 做得好的
- ...

### 做得不好的
- ...

### 根因模式
- 模式 A → 证据
- 模式 B → 证据

### 下期行动项
| 行动项 | 负责人角色 | 衡量指标 | 截止 |
|--------|------------|----------|------|
```