estat-api-data-search

e-Stat API Data Search

Use this skill to discover machine-readable e-Stat statistical tables.

Workflow

1. Read reference.md for endpoint behavior and response paths.
2. Decide whether the user needs downloadable files or API/DB-provided statistical tables. If file URLs are the target, use estat-file-search.
3. If the government statistics code is unknown, search the bundled government statistics code list with scripts/gov_stats_lookup.py.
4. Generate a small set of plausible searchWord values before calling the API. Prefer survey names, administrative terms,制度名, and likely table-title nouns over casual wording.
5. Search getStatsList with scripts/list.py, using statsCode and searchWord as the main levers.
6. Save the raw JSON response before narrowing or summarizing.
7. Inspect flattened candidates in e-Stat response order. Do not apply opaque scoring in the MVP.
8. Stop at statsDataId discovery. Metadata inspection and full data fetching are follow-up workflows.

Required environment

• ESTAT_APP_ID
  • Get one from https://www.e-stat.go.jp/mypage/user/preregister
  • Set it with export ESTAT_APP_ID="<your app id>", or pass --app-id for a one-off command.
  • If you only need to parse a saved response, use --from-fixture; no app ID is needed.

Scope

• Find candidate tables
• Prepare later data retrieval work

Operating rules

• Treat e-Stat API/DB search as a two-layer search space: government statistics code, then statistical table ID.
• Use government statistics code discovery and keyword search iteratively.
• searchKind is normally omitted; the API default is enough for this skill's table discovery workflow.
• statsNameList=Y is not used in MVP because it changes the response from table candidates to survey-name lists.
• @id is the statsDataId, but candidate output must include surrounding context because IDs can appear duplicated in unhelpful ways.
• Date fields such as UPDATED_DATE are display metadata, not authoritative filters.
• Prefer raw JSON plus CSV/JSONL candidate output over one-off jq pipelines for heavy responses.