AgentSkillsCN

notion-paper-retriever

利用低成本推理，从 Notion 中提取论文，助力材料表征学习与 RAE 项目。适用于用户要求从 Notion 加载或获取论文，并返回精炼的元数据列表时使用。

SKILL.md
--- frontmatter
name: notion-paper-retriever
description: Retrieve papers from Notion for the material representation learning + RAE project using low-cost reasoning. Use when the user asks to load or fetch papers from Notion and return concise metadata lists.

Notion Paper Retriever

Fetch project-relevant papers from Notion with minimal overhead and consistent filtering.

Quick start

  1. •Query the Paper data source.
  2. •Search within primary project relation filter by default.
  3. •If no match, fallback to full data source search.
  4. •Return concise fields in fixed order.

Workflow

0) If Notion MCP is not connected

  1. •Add Notion MCP:
    • •codex mcp add notion --url https://mcp.notion.com/mcp
  2. •Enable remote MCP client:
    • •set [features].rmcp_client = true in config.toml
    • •or run codex --enable rmcp_client
  3. •Login:
    • •codex mcp login notion

After login, restart Codex, then continue.

1) Use default targets

  • •Paper data source id: 7b347469-96b1-4b82-8c3f-9265f487cd4d
  • •Project page id: 412f4daa-a4ff-454f-929e-dab38d3718fe

2) Apply filters

  1. •Primary filter:
    • •Projects relation contains 412f4daa-a4ff-454f-929e-dab38d3718fe
    • •Run requested keyword search within this filtered subset first
  2. •Fallback filter:
    • •If primary filter returns no keyword match, rerun keyword search across the full Paper data source

3) Return fields (fixed order)

  1. •Name
  2. •Short Name
  3. •Year
  4. •Authors
  5. •URL
  6. •Status
  7. •Notion page URL

4) Reasoning mode

  • •Use low-cost reasoning by default for retrieval tasks.
  • •Avoid deep synthesis unless explicitly requested.

5) Failure handling

  • •If relation fields are missing, report the missing field and fallback to full data source keyword search.
  • •If results are large, return top N (if specified) or newest by year.