edinet-document-fetch

When to use

Use this skill when:

• You have a docID from edinet-document-search and need the actual document binary.
• The user wants to view a specific 有価証券報告書 PDF, 四半期報告書 XBRL, or 大量保有報告書 file.
• You're building a citation with a direct link to the source document.
• XBRL parsing or financial-statement extraction (manual / external tool) is the next step.

How to invoke

Call the kokai MCP server's get_edinet_document tool:

{
  "name": "get_edinet_document",
  "arguments": {
    "docID": "<8-char docID from search_edinet_documents>",
    "type": "2"
  }
}

type parameter (FSA 公式 §3-2-1)

Code	Format	Content
1	ZIP	提出本文書 + 監査報告書 (XBRL 含む)
2	PDF	提出本文書 + 監査報告書 (PDF 形式、最も読みやすい)
3	ZIP	代替書面・添付文書
4	ZIP	英文ファイル
5	ZIP	CSV (XBRL→CSV 変換版)

Size caps (Phase 1 limits)

• PDF (type=2): 50 MB
• ZIP (type=1/3/4/5): 100 MB

Oversize → tool returns edinet_document_too_large error with sizeBytes / maxBytes literal (Phase 2 で大型書類対応検討)。

Output structure

{
  "docID": "S1000001",
  "type": "2",
  "format": "pdf",
  "storageBucket": "edinet-documents",
  "storagePath": "S1000001/2/<timestamp>.pdf",
  "signedUrl": "https://...supabase.co/storage/v1/object/sign/edinet-documents/...",
  "sizeBytes": 1234567,
  "source_authority": "official",
  "attribution_text": "出典: 金融庁 EDINET",
  "boundary": {
    "signedUrlSource": "kokai_storage_proxy",
    "signedUrlTtlSeconds": 3600,
    "sourceAuthority": "edinet_official",
    "kokaiModification": "none",
    "canonicalSourceUrl": "https://disclosure2.edinet-fsa.go.jp"
  }
}

Important: signedUrl is a Kokai Storage proxy with a 1-hour TTL. The actual document content is unmodified from EDINET (kokaiModification: "none"). For canonical citation in long-lived documents, link to canonicalSourceUrl (EDINET 閲覧サイト) instead.

Boundary

• Output is signal / 確認材料 / context — NOT investment advice.
• The signed URL is for read access only (Kokai does not redistribute beyond the original requester).
• 投資判断 / 法的判断 / 税務判断 requires a certified financial advisor / 公認会計士 / 弁護士 / 税理士.
• EDINET disclosure 内容 はそのまま信頼できるが、Kokai は contents を解釈・要約していない (要約 / 推論は AI agent 側責任で source_authority: ai_summary / ai_estimate 4-layer authority strip に明示)。
• Public records only.