このファイルは
andon process renderがprocess/workstreams/watch.yamlから作ったもの。直すときは YAML を直して作り直す(手で直さない)。
目的: 担当の進みを見て、ズレたときだけ CoS が手を打つ
| きっかけ | 種類 | 中身 | 最初の手順 |
|---|---|---|---|
| patrol | 間隔 | 1 分ごと | look |
| estimate | 通知 | 通知 estimate |
read |
| stalled | 通知 | 通知 stalled |
read |
| error | 通知 | 通知 error |
read |
| finished | 通知 | 通知 finished |
read |
| wake | 出来事 | read |
flowchart LR
classDef machine fill:#dbeafe,stroke:#1d4ed8,color:#0b1b3a
classDef ai fill:#dcfce7,stroke:#15803d,color:#052e16
classDef person fill:#fef3c7,stroke:#b45309,color:#3b1d02
classDef event fill:#f3f4f6,stroke:#4b5563,color:#111827
t_patrol(("patrol")):::event
t_estimate(("estimate")):::event
t_stalled(("stalled")):::event
t_error(("error")):::event
t_finished(("finished")):::event
t_wake(("wake")):::event
subgraph lane_machine["機械"]
s_look("担当の端末と実働を見る"):::machine
g_look{"分かれ道"}
s_resume("死んだ担当を 1 回起動する"):::machine
end
subgraph lane_ai["AI"]
s_read("通知と計画を読む"):::ai
g_read{"分かれ道"}
s_act("担当に手を打つ"):::ai
s_tell("オーナーへ知らせる"):::ai
s_done("見るだけで終える"):::ai
end
t_patrol --> s_look
t_estimate --> s_read
t_stalled --> s_read
t_error --> s_read
t_finished --> s_read
t_wake --> s_read
s_look -->|"notification.estimate・notification.stalled・notification.error・notification.finished・hc send(「続けて」)"| g_look
g_look -->|"端末が Gone"| s_resume
g_look -->|"通知を積んだ"| s_read
s_resume -->|"起動できなかった"| s_read
s_read --> g_read
g_read -->|"手を打つ"| s_act
g_read -->|"オーナーへ知らせる"| s_tell
g_read -->|"見るだけ"| s_done
s_act -->|"hc send(担当へ)・kanban(分けた課題)"| fin
s_tell -->|"herdr notification・オーナーへの 1 問"| fin
s_done -->|"news.ack"| fin
fin(("終わり")):::event
| id | 手順 | やる者 | 入るもの | 出るもの | 次 |
|---|---|---|---|---|---|
look |
担当の端末と実働を見る | 機械 | issue.estimate・log.status・log.question・hc.state・hc.status_ms |
notification.estimate・notification.stalled・notification.error・notification.finished・hc send(「続けて」) |
resume(端末が Gone)・read(通知を積んだ) |
resume |
死んだ担当を 1 回起動する | 機械 | hc.state・issue.session |
hc(起動した端末)・notification.resumed・notification.finished |
read(起動できなかった) |
read |
通知と計画を読む | AI | notification.estimate・notification.stalled・notification.error・notification.finished・report.plan・hc.state・hc read の画面 |
act(手を打つ)・tell(オーナーへ知らせる)・done(見るだけ) |
|
act |
担当に手を打つ | AI | report.plan・hc read の画面 |
hc send(担当へ)・kanban(分けた課題) |
終わり |
tell |
オーナーへ知らせる | AI | hc read の画面 |
herdr notification・オーナーへの 1 問 |
終わり |
done |
見るだけで終える | AI | <news> |
news.ack |
終わり |
| id | コマンド | 記録 | 2 回起きてはいけない |
|---|---|---|---|
look |
andon dispatcher |
dispatcher/<slug>.json |
|
resume |
andon issue resume |
dispatcher/<slug>.json |
はい |
| id | どうやるか(SKILL の節・手引き) | 自分で決めてよいこと | オーナーに聞くこと |
|---|---|---|---|
read |
chief-of-staff/playbooks/watch.md |
途中報告を書かせる・分ける・起動し直す・延長する・続きを 1 回促す | |
act |
chief-of-staff/playbooks/watch.md |
送る文・分けた課題の範囲 | |
tell |
chief-of-staff/playbooks/watch.md・ask-one-by-one |
知らせる文 | 死んだまま起きない担当を新しいセッションで起動し直すか(andon issue resume --fresh)(business) |
done |
chief-of-staff/playbooks/watch.md |
何もしない |
見るもの(手順の入るものの和): issue.estimate・log.status・log.question・hc.state・hc.status_ms・issue.session・notification.estimate・notification.stalled・notification.error・notification.finished・report.plan・hc read の画面・<news>
残すもの(手順の出るものの和): notification.estimate・notification.stalled・notification.error・notification.finished・hc send(「続けて」)・hc(起動した端末)・notification.resumed・hc send(担当へ)・kanban(分けた課題)・herdr notification・オーナーへの 1 問・news.ack
flow.wait.todo・flow.wait.queued・flow.wait.doing・flow.wait.review