AgentSkillsCN

logseq-daily-summary

在 LogSeq 日记中每日撰写并追加不超过 300 字的中文摘要。当用户希望总结今日 LogSeq 日记、分析某一天的日记内容，或把每日摘要重新写回 LogSeq 时，可使用此技能。

SKILL.md
--- frontmatter
name: logseq-daily-summary
description: Create and append a <=300 Chinese-character daily summary into LogSeq journals. Use when the user asks to summarize today’s LogSeq journal, analyze a specific day’s journal, or write a daily summary back into LogSeq.

Logseq Daily Summary

Overview

Summarize a LogSeq daily journal (today by default, or a specified log) and append a concise <=300 Chinese-character paragraph back into that day’s journal page.

Workflow (LogSeq journal -> usage stats -> summary -> write back)

  1. •Determine the target journal page.
  • •If the user specifies a log/date/page name, use that.
  • •Otherwise use today’s date based on local time. Use list_pages (include journals) and find a page whose title starts with YYYY-MM-DD. LogSeq journal titles may be YYYY-MM-DD DayName (English), YYYY-MM-DD 星期X (Chinese), or just YYYY-MM-DD.
  • •If multiple pages match, prefer the one marked [journal], then the one with the longest non-empty content (via get_page_content), then the English weekday variant.
  • •If no page matches the date, use search to confirm. If still missing, ask the user for the exact page name before creating a new page.
  1. •Read the journal content.
  • •Use get_page_content to fetch the page content.
  • •If the page does not exist, create it (via create_page or update_page) and note it was empty before summarizing.
  1. •Collect today's Codex CLI usage stats and append a usage block.
  • •Use the fixed script: scripts/codex_usage_stats.py.
  • •Example: python3 /Users/wushitao/.codex/skills/logseq-daily-summary/scripts/codex_usage_stats.py --date YYYY-MM-DD.
  • •The script reads ~/.codex/sessions/YYYY/MM/DD/*.jsonl, aggregates the last token_count.total_token_usage per session file, and extracts the local time range for the day (min/max timestamps).
  • •Append the script output as its own LogSeq block before the summary block, using the exact format below (no leading "- "):
code
🤖 Codex CLI 消耗（自动统计）
  ⏱️ 时间段: YYYY-MM-DD HH:MM:SS–HH:MM:SS (+0800)
  🧾 用量: N tokens
  🧠 Token: 输入 A / 输出 B / 缓存输入 C / 推理 D
  1. •Write the summary.
  • •Produce a single paragraph in Chinese, 300 characters or fewer (do not count the usage block).
  • •Focus on “what I was thinking about / investigating today,” derived from the journal content.
  • •Avoid lists in the summary paragraph; keep it clean and readable.
  1. •Append to the same journal page.
  • •Use update_page to append the usage block (if not already present), then use a second update_page call to append the summary block 今日总结: ... (no leading "- ").
  • •Ensure the usage block and summary block are separate top-level blocks and never nested, and do not prefix lines with "- ".
  • •Do not overwrite existing content.

Output rules

  • •Summary length: <=300 Chinese characters (usage block not counted).
  • •One paragraph only.
  • •Keep it faithful to the journal content; no fabrication.
  • •Always include the Codex CLI usage block in the required format (from the fixed script).
  • •Usage block and summary block must be separate top-level blocks.

Tools

Use MCP LogSeq tools:

  • •search or list_pages to locate the journal page
  • •get_page_content to read
  • •update_page to append the summary
  • •create_page only if necessary