groove-utilities-memory-retrospective

Compare original and translation side by side

🇺🇸

Original

English
🇨🇳

Translation

Chinese

groove-utilities-memory-retrospective

groove-utilities-memory-retrospective

Use $ARGUMENTS to specify period:
week
(default),
month
, or
all
.
使用$ARGUMENTS指定统计周期:
week
(默认)、
month
all

Outcome

输出结果

A period retrospective is output to the conversation: session rating trend, recurring mistake patterns, top learnings captured, and a short workflow health summary. No files are written.
会话中会输出指定周期的复盘内容:会话评分趋势、重复错误模式、捕获的核心经验收获,以及简短的工作流健康度总结,不会写入任何文件。

Acceptance Criteria

验收标准

  • Session ratings trend shown (if
    learned/signals.md
    has data)
  • Mistake patterns summarised (from resolved incidents in task backend)
  • Top learnings listed by topic (from
    learned/*.md
    dated entries)
  • Workflow health summary: one sentence on overall trend
  • All output goes to conversation only
  • 展示会话评分趋势(若
    learned/signals.md
    存在数据)
  • 总结错误模式(取自任务后端中已解决的事件)
  • 按主题列出核心经验收获(取自
    learned/*.md
    下带日期的条目)
  • 工作流健康度总结:一句话说明整体趋势
  • 所有内容仅输出到会话中

Steps

执行步骤

  1. Read
    memory:
    from
    .groove/index.md
    (default:
    .groove/memory/
    )
  2. Determine date range from $ARGUMENTS:
    • week
      : last 7 days (today - 7)
    • month
      : last 30 days
    • all
      : all available data
    • Default if absent:
      week
  3. Session ratings — read
    <memory>/learned/signals.md
    if it exists:
    • Parse table rows matching the date range:
      | YYYY-MM-DD | N/5 | context |
    • Compute: count, average rating, trend (last 3 vs first 3 if enough data)
    • If no data: note "No session ratings recorded yet"
  4. Mistakes — query the task backend for resolved incidents under "Groove Memory" → "Mistakes":
    • beans list --parent <mistakes-epic-id> -t bug -s completed
      for resolved;
      -s in-progress
      for open
    • Group by title keywords to identify recurring patterns
    • Count total resolved vs open incidents
    • If no task backend or no data: note "No incidents recorded yet"
  5. Learnings — glob
    <memory>/learned/*.md
    (exclude
    signals.md
    ):
    • For each file: read and collect dated entries (
      ## YYYY-MM-DD
      headings) within the date range
    • Group by topic (filename stem)
    • Count entries per topic; show the most recent bullet per topic as a sample
    • If no files: note "No learnings captured yet"
  6. Git activity — run
    git log --oneline --since="<start-date>"
    to count commits in the period as a proxy for work volume
  7. Output retrospective to conversation in this format:
undefined
  1. .groove/index.md
    读取
    memory:
    路径(默认值:
    .groove/memory/
  2. 根据$ARGUMENTS确定日期范围:
    • week
      :最近7天(今日往前推7天)
    • month
      :最近30天
    • all
      :所有可用数据
    • 无参数时默认取
      week
  3. 会话评分 —— 若存在
    <memory>/learned/signals.md
    则读取该文件:
    • 解析符合日期范围的表格行:
      | YYYY-MM-DD | N/5 | context |
    • 计算:评分次数、平均评分、趋势(数据量足够时对比前3条和后3条)
    • 无数据时备注:"No session ratings recorded yet"
  4. 错误统计 —— 向任务后端查询「Groove Memory」→「Mistakes」分类下已解决的事件:
    • 已解决事件查询命令:
      beans list --parent <mistakes-epic-id> -t bug -s completed
      ;待解决事件用
      -s in-progress
    • 按标题关键词分组识别重复出现的模式
    • 统计已解决/待解决事件总数
    • 无任务后端或无数据时备注:"No incidents recorded yet"
  5. 经验收获 —— 匹配
    <memory>/learned/*.md
    下的所有文件(排除
    signals.md
    ):
    • 读取每个文件,收集日期范围内带日期的条目(
      ## YYYY-MM-DD
      标题)
    • 按主题分组(主题为文件名前缀)
    • 统计每个主题的条目数量,展示每个主题最新的一条要点作为示例
    • 无文件时备注:"No learnings captured yet"
  6. Git活动统计 —— 执行
    git log --oneline --since="<start-date>"
    统计周期内的提交次数,作为工作量参考
  7. 按以下格式将复盘内容输出到会话中:
undefined

Groove Retrospective — <period> (<start-date> to <today>)

Groove Retrospective — <period> (<start-date> to <today>)

Session Ratings

Session Ratings

<count> sessions rated | avg: <N>/5 | trend: <up/down/flat or n/a> <sparkline of ratings if ≥3 data points, e.g. ▂▄█▆▇>
<count> sessions rated | avg: <N>/5 | trend: <up/down/flat or n/a> <sparkline of ratings if ≥3 data points, e.g. ▂▄█▆▇>

Mistakes

Mistakes

<N> incidents resolved | <N> open Top patterns:
  • <root cause pattern> (N occurrences)
  • ... (No recurring patterns) if all unique
<N> incidents resolved | <N> open Top patterns:
  • <root cause pattern> (N occurrences)
  • ... (No recurring patterns) if all unique

Learnings

Learnings

<N total entries across <M> topics>
  • patterns (<N> entries): <most recent bullet>
  • anti-patterns (<N> entries): <most recent bullet>
  • ...
<N total entries across <M> topics>
  • patterns (<N> entries): <most recent bullet>
  • anti-patterns (<N> entries): <most recent bullet>
  • ...

Workflow Health

Workflow Health

<git commits in period> commits | <1-sentence summary of overall trend> e.g. "Stable week — ratings above 4, one incident (resolved), 3 new patterns captured."
undefined
<git commits in period> commits | <1-sentence summary of overall trend> e.g. "Stable week — ratings above 4, one incident (resolved), 3 new patterns captured."
undefined

Sparkline generation

趋势迷你图生成

Use block characters to render a simple ratings sparkline:
  • 1/5 → ▁, 2/5 → ▂, 3/5 → ▄, 4/5 → ▆, 5/5 → █
  • Only show if ≥ 3 ratings in the period
使用块字符渲染简单的评分趋势迷你图:
  • 1/5 → ▁, 2/5 → ▂, 3/5 → ▄, 4/5 → ▆, 5/5 → █
  • 仅当周期内评分数量≥3时展示迷你图

Constraints

约束条件

  • Read only — do not write to any file
  • If a data source file is absent, skip that section gracefully with a one-line note
  • Parse table rows defensively: skip malformed rows rather than erroring
  • Date filtering: match rows where the
    YYYY-MM-DD
    date falls within the range; skip rows without a valid date
  • Output to conversation only — never write to AGENTS.md or memory files
  • 只读操作 —— 不得写入任何文件
  • 若数据源文件不存在,友好跳过对应模块并展示一行备注
  • 容错解析表格行:跳过格式错误的行而不是抛出错误
  • 日期过滤:仅匹配
    YYYY-MM-DD
    格式日期在范围内的行,跳过无有效日期的行
  • 仅输出到会话 —— 禁止写入AGENTS.md或内存文件