Daily-report revival investigation and v1 implementation

過去レポートのView/ソース規律バックフィルで生成したView。

Generated: 2026-05-10T22:27:00+09:00

What existed before

Legacy secretary daily/status reporting lived mostly under /Users/aiharataketo/secretary-state/:

• standup.sh sent Telegram standups for morning|noon|evening.

- Header: ☀️ 朝会 / 🍱 昼会 / 🌙 夜会.

- Sections: 🔴 要対応, 🔵 作業中, ✅ 最近の完了, 📦 BACKLOG count.

- Source: jobs.db statuses (WAIT/BLOCKED/PENDING, SENT/IN_PROGRESS, recent DONE).

• report-done.sh generated verbose completion reports from jobs.db, verification_summary, done_when.

- Four sections: 背景 / やったこと / あなたへの影響 / 次のアクション.

• report-done-v2.sh replaced the verbose completion report with a short 3 lines + URL format.

- The file comment explicitly says old bracketed headings and multi-level bullets were removed, and details were pushed into the dashboard/View.

• gen-completion-summary.sh generated plan-viewer completion pages and updated jobs.plan_viewer_url.

Owner feedback found

Direct daily-report-specific feedback in OpenClaw sessions was one current request:

> 前の秘書でやってたdaily-reportを復活。それに対する俺のコメントや文句を拾って、継続的改善をする仕組みを整えて欲しい

Broader feedback relevant to reports/status:

• Do not rely on short-term memory; record task state durably.
• Separate “investigated” from “done”; done needs execution/verification/evidence.
• Safe internal work should proceed without repeated confirmation.
• Avoid over-reportish/Jira/admin tone in chat; put detail behind reports/View links.
• When asked to check with another agent/source, do not jump to a conclusion first.

v1 implementation

Added scripts/daily_report.py:

• Reads Markdown task ledger from projects/task-queue-visibility/tasks/*.md.
• Reads OpenClaw runtime tasks with openclaw tasks list --json when available.
• Scans same-day local session logs for notable owner requests/corrections.
• Writes a local Markdown daily report under projects/task-queue-visibility/reports/daily/YYYY-MM-DD.md.
• Never sends Telegram/email and never creates cron.

Generated v1 report:

• projects/task-queue-visibility/reports/daily/2026-05-10.md

Added scripts/owner_feedback_miner.py:

• Mines recent local sessions for owner complaints/corrections.
• Categories include daily_report_feedback, too_verbose_or_reportish, too_short_or_missing_context, premature_conclusion_or_overclaim, autonomy_preference, and task_capture_gap.
• Writes report-first Markdown under projects/task-queue-visibility/reports/.
• Does not mutate persona/config/runtime.

Generated feedback reports:

• projects/task-queue-visibility/reports/owner-feedback-miner-20260510-222429.md
• projects/task-queue-visibility/reports/owner-feedback-miner-daily_report_feedback-20260510-222429.md

Validation

Passed:

python3 -m py_compile scripts/daily_report.py scripts/owner_feedback_miner.py scripts/test_daily_report_and_feedback.py
python3 scripts/test_daily_report_and_feedback.py
# Ran 4 tests ... OK
python3 scripts/daily_report.py --date 2026-05-10 --json
python3 scripts/owner_feedback_miner.py --limit-files 20 --max-hits 80 --json
python3 scripts/owner_feedback_miner.py --focus daily_report_feedback --limit-files 40 --max-hits 80 --json
python3 scripts/task_ledger.py lint
# TASK_LEDGER_OK 28 tasks

Approval boundary / remaining work

Not done intentionally:

• No Telegram delivery cron.
• No persistent schedule.
• No external send.

To enable actual daily delivery, the owner should approve:

1. schedule, e.g. daily 21:30 JST or morning/noon/evening,
2. destination, e.g. Telegram DM 7789180125,
3. exact message style/body preview or “send report link only”.