القائمة
Morning paper reading workflow. Triggers on "start my day", "今天看什么", "morning digest", "每日推荐".
التثبيت باستخدام Codex أو Claude انسخ هذا Prompt والصقه في Codex أو Claude أو مساعد آخر ليراجع صفحة Skill ويثبّتها لك.
Citation chain exploration. Triggers on "引用链", "谁引了", "citation", "参考文献", "这篇引了谁".
Deep analysis of a single paper. Triggers on "分析这篇", "deep dive", "这篇论文讲了什么", paper ID or arXiv ID input.
Literature survey and review workflow. Triggers on "survey", "综述", "这个方向有什么", "literature review", "related work".
AI research paper intelligence — router skill that detects user intent and delegates to the appropriate workflow. Triggers on paper, arxiv, digest, research, 论文, arXiv IDs, ML method names, or academic research questions.
Batch paper screening and classification. Triggers on "筛一下", "triage", "哪些值得看", "帮我筛", "batch screen".
Research trend analysis and insight generation. Triggers on "趋势", "trend", "这个方向火不火", "research insight", "方向分析".
استنادا إلى تصنيف SOC المهني
| name | daily-reading |
| description | Morning paper reading workflow. Triggers on "start my day", "今天看什么", "morning digest", "每日推荐". |
Call paper_morning_brief(days=1) — this single tool does context recovery + collect + digest + auto-mark in one call
Call paper_watch_digest() — check for watchlist updates
Present as structured tables:
今日推荐 (N 篇): | # | 标题 | 评分 | 关键词 | 一句话总结 |
Watchlist 更新 (if any): | 跟踪项 | 类型 | 新论文数 | 最相关论文 |
结论与建议: 今日最值得关注的方向和论文，建议阅读顺序
Auto-save: call paper_save_report(report_type="daily_digest", content=<digest markdown>, filename="{YYYY-MM-DD}.md") using the daily-digest-template. Tell user the saved path.
[CONTEXT-AWARE FORK] — Based on the digest result, suggest next steps:
If digest has ≥1 high-confidence paper: "今日摘要已保存至 {path}。\n💡 下一步建议：\n1. 深入分析 [{top_paper_title}]（{score}分，{reason}）\n2. 先筛选一轮，确认哪些值得读\n（说编号或告诉我你想做什么）"
If digest is empty: "今天没有新的高相关论文。\n💡 下一步建议：\n1. 扫描一下 [{user_primary_topic}] 的最新进展\n2. 看看 watchlist 有没有更新\n（说编号或告诉我你想做什么）"
If watchlist has updates: Append: "另外，[{watch_item}] 有 {N} 篇新论文，要看看吗？"
Hand off to the deep-dive skill with the selected paper ID. Carry the paper's score, topics, and recommendation_reason as context.
The paper_morning_brief response includes a mode field:
"workspace": top picks were auto-marked as to_read.
auto_marked > 0 AND this is the user's first time seeing auto-marking (no prior reading stats), explain: "（首次自动标记）我把 N 篇高相关论文自动标记为'待读'，方便你追踪阅读进度。这个行为可以通过切换到 lightweight 模式关闭。""lightweight": just present the digest, no status talk.