generate-standup

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

Generate Daily Standup

生成每日站会内容

Produces a daily standup post from live GitHub activity and optionally posts it to the

mitodl/hq

Check-ins discussion.

Requires:

gh

(authenticated) and

jq

基于实时GitHub活动生成每日站会帖子，并可选择发布至

mitodl/hq

的Check-ins讨论区。

**依赖：**已认证的

gh

和

jq

。

Step 1 — Fetch GitHub context

步骤1 — 获取GitHub上下文

Run the bundled context script before asking any questions:

bash

bash skills/process/generate-standup/scripts/get-standup-context.sh [-t YYYY-MM-DD] [-o org1,org2]

Flag	Description	Default
`-t`	"Today" date ( `YYYY-MM-DD` )	today (UTC)
`-o`	Comma-separated orgs to search	`mitodl,openedx`

The script outputs a JSON object:

json

{
  "meta": { "username", "today", "yesterday", "tomorrow", "since" },
  "checkin_discussion": { "id", "number", "title", "url", "createdAt" },
  "prs_authored":    [...],
  "prs_reviewed":    [...],
  "issues":          [...],
  "rfc_discussions": [...]
}

```
meta.yesterday
```
is the previous weekday (Friday if today is Monday).
```
meta.tomorrow
```
is the next weekday (Monday if today is Friday).
```
meta.since
```
is midnight UTC on
```
meta.yesterday
```
— the fetch window start.
```
checkin_discussion
```
is the most recent Check-ins discussion in
```
mitodl/hq
```
. Keep its
```
id
```
(GraphQL node ID) and
```
url
```
for Steps 4–5.
Do not infer or fabricate activity beyond what the script returns.

在询问任何问题之前，先运行内置的上下文脚本：

bash

bash skills/process/generate-standup/scripts/get-standup-context.sh [-t YYYY-MM-DD] [-o org1,org2]

参数	描述	默认值
`-t`	"今日"日期（格式 `YYYY-MM-DD` ）	今日（UTC时间）
`-o`	逗号分隔的要搜索的组织	`mitodl,openedx`

脚本会输出一个JSON对象：

json

{
  "meta": { "username", "today", "yesterday", "tomorrow", "since" },
  "checkin_discussion": { "id", "number", "title", "url", "createdAt" },
  "prs_authored":    [...],
  "prs_reviewed":    [...],
  "issues":          [...],
  "rfc_discussions": [...]
}

```
meta.yesterday
```
为前一个工作日（若今日为周一，则为上周五）。
```
meta.tomorrow
```
为下一个工作日（若今日为周五，则为下周一）。
```
meta.since
```
为
```
meta.yesterday
```
的UTC午夜时间——即数据获取窗口的起始时间。
```
checkin_discussion
```
是
```
mitodl/hq
```
中最新的Check-ins讨论区。请保留其
```
id
```
（GraphQL节点ID）和
```
url
```
用于步骤4-5。
请勿推断或编造脚本返回内容之外的活动信息。

Step 1b — Query agent session history

步骤1b — 查询Agent会话历史

Using the

sql

tool (

database: "session_store"

), fetch agent sessions active since

meta.since

sql

SELECT
  s.id,
  s.repository,
  s.branch,
  s.summary,
  s.created_at,
  s.updated_at,
  c.title        AS checkpoint_title,
  c.overview     AS checkpoint_overview,
  c.work_done    AS checkpoint_work_done
FROM sessions s
LEFT JOIN checkpoints c ON c.session_id = s.id
WHERE s.updated_at >= '<meta.since>'
ORDER BY s.updated_at DESC

For sessions with no checkpoints, fetch the first user turn as a fallback:

sql

SELECT s.id, s.repository, s.branch, t.user_message
FROM sessions s
JOIN turns t ON t.session_id = s.id AND t.turn_index = 0
WHERE s.updated_at >= '<meta.since>'
  AND NOT EXISTS (SELECT 1 FROM checkpoints c WHERE c.session_id = s.id)
ORDER BY s.updated_at DESC

Summarization rules:

Evidence available	Action
Checkpoint with `work_done` / `overview`	Use as session summary
No checkpoint; has repo + branch + concrete first turn	Derive brief summary from repo/branch + turn intent
No checkpoint; NULL repo or trivial/meta prompt	Skip
Session is for generating this standup	Skip

Store the resulting list of session summaries; use in Steps 3–4 to enrich GitHub-derived bullets and fill in non-GitHub work.

使用

sql

工具（

database: "session_store"

）获取自

meta.since

以来活跃的Agent会话：

sql

SELECT
  s.id,
  s.repository,
  s.branch,
  s.summary,
  s.created_at,
  s.updated_at,
  c.title        AS checkpoint_title,
  c.overview     AS checkpoint_overview,
  c.work_done    AS checkpoint_work_done
FROM sessions s
LEFT JOIN checkpoints c ON c.session_id = s.id
WHERE s.updated_at >= '<meta.since>'
ORDER BY s.updated_at DESC

对于无检查点的会话，获取用户的第一条消息作为备选：

sql

SELECT s.id, s.repository, s.branch, t.user_message
FROM sessions s
JOIN turns t ON t.session_id = s.id AND t.turn_index = 0
WHERE s.updated_at >= '<meta.since>'
  AND NOT EXISTS (SELECT 1 FROM checkpoints c WHERE c.session_id = s.id)
ORDER BY s.updated_at DESC

总结规则：

可用证据	操作
包含 `work_done` / `overview` 的检查点	用作会话总结
无检查点；包含仓库+分支+明确的第一条消息	从仓库/分支和消息意图生成简短总结
无检查点；仓库为NULL或提示无实际内容	跳过
会话用于生成本次站会内容	跳过

保存生成的会话总结列表；在步骤3-4中用于丰富GitHub衍生的项目符号，并补充GitHub外的工作内容。

Step 2 — Ask clarifying questions

步骤2 — 询问澄清问题

Use a single
ask_user
call with all fields at once.

First, identify session-only work from Step 1b (sessions with

repository: null

or whose repository doesn't appear in

prs_authored

). Format them as a short suggestion list for the

off_github

field description.

The most important field is

timing

— it controls the section headers and which date's activity is treated as "done" work:

json

{
  "timing": {
    "type": "string",
    "title": "When are you posting?",
    "enum": ["EOD — reporting today's work (today/tomorrow headers)",
             "BOD — reporting yesterday's work (yesterday/today headers)"],
    "description": "EOD: post at end of your work day; yesterday section covers today's date (meta.today). BOD: post at start of your work day; yesterday section covers meta.yesterday."
  },
  "blockers": {
    "type": "string",
    "title": "Blockers",
    "description": "Are you blocked on anything? Include a link and the @handle of whoever needs to unblock you. Leave blank if none."
  },
  "announcements": {
    "type": "string",
    "title": "Announcements",
    "description": "Anything to announce not in GitHub? (OOO, special review requests, schedule changes, etc.) Leave blank if none."
  },
  "off_github": {
    "type": "string",
    "title": "Off-GitHub work",
    "description": "Meetings, planning, research, design, or other work that won't appear in GitHub. Leave blank if none.\n\nPossible session-only work detected:\n<bullet list of session-only summaries, or 'none detected'>"
  }
}

使用单次

ask_user

调用，一次性获取所有字段信息。

首先，从步骤1b中识别仅会话中的工作（

repository

为NULL或仓库未出现在

prs_authored

中的会话）。将其格式化为

off_github

字段描述的简短建议列表。

最重要的字段是

timing

——它控制章节标题以及将哪一天的活动视为“已完成”工作：

json

{
  "timing": {
    "type": "string",
    "title": "你将在何时发布？",
    "enum": ["EOD — 汇报今日工作（今日/明日标题）",
             "BOD — 汇报昨日工作（昨日/今日标题）"],
    "description": "EOD：在工作日结束时发布；昨日章节对应今日日期（meta.today）。BOD：在工作日开始时发布；昨日章节对应meta.yesterday。"
  },
  "blockers": {
    "type": "string",
    "title": "阻塞事项",
    "description": "你是否遇到任何阻塞？请包含链接和需要协助解除阻塞的人员@用户名。若无则留空。"
  },
  "announcements": {
    "type": "string",
    "title": "公告",
    "description": "是否有未在GitHub上体现的公告？（休假、特殊评审请求、日程变更等）若无则留空。"
  },
  "off_github": {
    "type": "string",
    "title": "GitHub外的工作",
    "description": "会议、规划、研究、设计或其他未在GitHub上体现的工作。若无则留空。\n\n检测到的可能仅会话工作：\n<仅会话总结的项目符号列表，或'未检测到'>"
  }
}

Step 3 — Classify activity

步骤3 — 分类活动

From the user's

timing

answer, determine:

Timing

Timing	`report_date` (done work)	`planned_date` (next work)	Past header	Future header
EOD	`meta.today`	`meta.tomorrow`	`What did I work on today?`	`What am I working on tomorrow?`
BOD	`meta.yesterday`	`meta.today`	`What did I work on yesterday?`	`What am I working on today?`

report_date

(done work)

planned_date

(next work)

Past header

Future header

EOD

meta.today

meta.tomorrow

What did I work on today?

What am I working on tomorrow?

BOD

meta.yesterday

meta.today

What did I work on yesterday?

What am I working on today?

Bucketing rules:

Done (past section): Any PR or issue with
```
updatedAt
```
or
```
mergedAt
```
on
```
report_date
```
. Include both merged and still-open items that were actively worked on that day.
Planned (future section): Open PRs and issues the user is continuing, plus anything explicitly stated in user answers. Omit items with no
```
updatedAt
```
since
```
meta.since
```
(stale).
Announcements: PRs authored by the user that are still open and need human review (exclude bots: Copilot, Gemini, Renovate, Dependabot, Sentry). Also include RFC discussions created today, blockers, and OOO info.
Deduplication: A PR in both
```
prs_authored
```
and
```
prs_reviewed
```
→ list once under the most relevant bucket.

Incorporating agent sessions:

If a session maps to a PR/issue already in the GitHub data, enrich that bullet with context from the session summary — do not create a duplicate.
If a session represents work with no GitHub artifact, add it as its own bullet under done or planned based on
```
updated_at
```
vs
```
report_date
```
.

根据用户的

timing

回答，确定：

时间选择

时间选择	`report_date` （已完成工作）	`planned_date` （后续工作）	过去章节标题	未来章节标题
EOD	`meta.today`	`meta.tomorrow`	`我今日完成了哪些工作？`	`我明日将开展哪些工作？`
BOD	`meta.yesterday`	`meta.today`	`我昨日完成了哪些工作？`	`我今日将开展哪些工作？`

report_date

（已完成工作）

planned_date

（后续工作）

过去章节标题

未来章节标题

EOD

meta.today

meta.tomorrow

我今日完成了哪些工作？

我明日将开展哪些工作？

BOD

meta.yesterday

meta.today

我昨日完成了哪些工作？

我今日将开展哪些工作？

分类规则：

**已完成（过去章节）：**任何在
```
report_date
```
有
```
updatedAt
```
或
```
mergedAt
```
记录的PR或Issue。包含已合并和仍在进行中且当日有活跃操作的项目。
**计划中（未来章节）：**用户将继续推进的开放PR和Issue，加上用户回答中明确提及的事项。排除自
```
meta.since
```
以来无
```
updatedAt
```
记录的陈旧项目。
**公告：**用户发起的仍开放且需要人工评审的PR（排除机器人：Copilot、Gemini、Renovate、Dependabot、Sentry）。还包括今日创建的RFC讨论、阻塞事项和休假信息。
**去重：**同时出现在
```
prs_authored
```
和
```
prs_reviewed
```
中的PR → 归入最相关的分类一次。

整合Agent会话：

如果会话对应GitHub数据中已有的PR/Issue，用会话总结的上下文丰富该项目符号——不要创建重复项。
如果会话代表无GitHub工件的工作，根据
```
updated_at
```
与
```
report_date
```
的关系，将其作为单独的项目符号添加到已完成或计划中章节。

Step 4 — Render the standup

步骤4 — 渲染站会内容

Use

meta.username

(the GitHub login) as the name.

markdown

<Display Name>

> Standup announcements

- <item>

> <past header>

- <item>

> <future header>

- <item>

Formatting rules:

Empty sections: write
```
- None
```
, never omit the section header.
Blockers go in announcements as a bullet; tag with
```
@handle
```
and link.
Links: raw GitHub URLs are fine; markdown
```
[text](url)
```
formatting is also fine — match what feels natural for the content. Don't force one style.
Level of detail: match what the data supports. If a PR/issue title is self-explanatory, a bare link is sufficient. Add a brief description only when context genuinely helps (e.g., the PR title doesn't convey purpose, or the work involved investigation/discussion not captured in a link).
Do not impose narrative style: some people post links; some post prose; both are correct. Let the available data guide the output.
Do not group PRs across repos into parent bullets unless the user explicitly works across many repos on the same thing and grouping is clearly cleaner — default to separate bullets.
Planned section should reflect what's actually next, not a mechanical list of every open PR. Omit items the user is clearly done with.

使用

meta.username

（GitHub登录名）作为显示名称。

markdown

<显示名称>

> 站会公告

- <项目>

> <过去章节标题>

- <项目>

> <未来章节标题>

- <项目>

格式规则：

**空章节：**填写
```
- None
```
，切勿省略章节标题。
阻塞事项作为项目符号放在公告中；标记
```
@用户名
```
并添加链接。
**链接：**原始GitHub URL即可；也可使用Markdown的
```
[文本](链接)
```
格式——根据内容选择自然的样式，无需强制统一。
**详细程度：**匹配数据支持的内容。如果PR/Issue标题已清晰说明，仅需链接即可。仅当上下文确实有帮助时（例如PR标题无法传达目的，或工作涉及未在链接中体现的调查/讨论）才添加简短描述。
**不要强加叙事风格：**有些人只发链接，有些人发段落文字；两种方式都合理。根据可用数据决定输出形式。
不要跨仓库分组PR到父项目符号下，除非用户明确跨多个仓库处理同一事务且分组明显更清晰——默认使用单独的项目符号。
计划章节应反映实际后续工作，而非机械列出所有开放PR。排除用户显然已完成的项目。

Step 5 — Confirm and post

步骤5 — 确认并发布

Display the rendered standup, then use

ask_user

to confirm:

json

{
  "action": {
    "type": "string",
    "title": "Post this standup?",
    "enum": ["Post it", "Edit first", "Cancel"],
    "description": "Post as a comment on <title> (<checkin_discussion.url>), make edits, or cancel."
  }
}

Do not post unless the user selects "Post it".

On confirmation, post using the bundled script:

bash

echo "<rendered standup>" \
  | bash skills/process/generate-standup/scripts/post-standup-comment.sh \
      -d "<checkin_discussion.id>"

The script prints the comment URL on success.

显示渲染后的站会内容，然后使用

ask_user

获取确认：

json

{
  "action": {
    "type": "string",
    "title": "发布此站会内容？",
    "enum": ["发布", "先编辑", "取消"],
    "description": "发布为<title>的评论（<checkin_discussion.url>），进行编辑，或取消操作。"
  }
}

仅当用户选择“发布”时才可发布。

确认后，使用内置脚本发布：

bash

echo "<渲染后的站会内容>" \
  | bash skills/process/generate-standup/scripts/post-standup-comment.sh \
      -d "<checkin_discussion.id>"

脚本成功执行后会打印评论链接。

Example output (EOD, link-primary style)

示例输出（EOD，以链接为主风格）

markdown

Anna G

> Standup announcements

- https://github.com/mitodl/mitxonline/pull/3600 needs review

> What did I work on today?

- worked on https://github.com/mitodl/mitxonline/pull/3600
- updated UI and fixed tests https://github.com/mitodl/mit-learn/pull/3346, received review

> What am I working on tomorrow?

- address feedback https://github.com/mitodl/mit-learn/pull/3346
- resolve https://github.com/mitodl/hq/issues/11440

markdown

Anna G

> 站会公告

- https://github.com/mitodl/mitxonline/pull/3600 需要评审

> 我今日完成了哪些工作？

- 处理了 https://github.com/mitodl/mitxonline/pull/3600
- 更新UI并修复测试 https://github.com/mitodl/mit-learn/pull/3346，已收到评审意见

> 我明日将开展哪些工作？

- 处理 https://github.com/mitodl/mit-learn/pull/3346 的反馈
- 解决 https://github.com/mitodl/hq/issues/11440

Example output (BOD, narrative style)

示例输出（BOD，叙事风格）

markdown

Tobias Macey

> Standup announcements

- PRs needing review:
  - [ol-infrastructure #4659: add Archive/Deep Archive access tier support to OLBucket](https://github.com/mitodl/ol-infrastructure/pull/4659)
  - [ol-data-platform #2238: automate Iceberg table maintenance across the lakehouse](https://github.com/mitodl/ol-data-platform/pull/2238)

> What did I work on yesterday?

- Worked on addressing the hanging open issue for Dagster assets using Polars to read Iceberg tables
- Opened https://github.com/mitodl/ol-infrastructure/pull/4659 for S3 cost optimization

> What am I working on today?

- Finish fixing the Polars/Iceberg hang in Dagster
- Test the Concourse release workflow end to end
- Wrap up self assessment

markdown

Tobias Macey

> 站会公告

- 需要评审的PR：
  - [ol-infrastructure #4659: 为OLBucket添加归档/深度归档访问层级支持](https://github.com/mitodl/ol-infrastructure/pull/4659)
  - [ol-data-platform #2238: 自动化湖仓中Iceberg表的维护](https://github.com/mitodl/ol-data-platform/pull/2238)

> 我昨日完成了哪些工作？

- 处理使用Polars读取Iceberg表的Dagster资产悬而未决的问题
- 为S3成本优化提交了 https://github.com/mitodl/ol-infrastructure/pull/4659

> 我今日将开展哪些工作？

- 完成Dagster中Polars/Iceberg挂起问题的修复
- 端到端测试Concourse发布流程
- 完成自我评估

Example output (ambiguous timing, hybrid style)

示例输出（时间模糊，混合风格）

markdown

Sar

> Standup announcements

- None

> What did I work on yesterday/today?

- Wrote and deployed https://github.com/mitodl/ol-infrastructure/pull/4658
- Continued investigating SCIM sync failures — updates are reaching Keycloak
  logs but not propagating to Learn/MITx Online; restarting Keycloak temporarily
  restores sync, root cause still unknown

> What am I working on today/tomorrow?

- Continue digging into the SCIM update issue

See context script for the GitHub data-fetching implementation and post script for the comment posting implementation.

markdown

Sar

> 站会公告

- None

> 我昨日/今日完成了哪些工作？

- 编写并部署了 https://github.com/mitodl/ol-infrastructure/pull/4658
- 继续调查SCIM同步失败问题——更新已到达Keycloak日志，但未同步到Learn/MITx Online；重启Keycloak可临时恢复同步，根本原因仍未知

> 我今日/明日将开展哪些工作？

- 继续深入排查SCIM更新问题

查看上下文脚本了解GitHub数据获取实现，查看发布脚本了解评论发布实现。