For agents

Use this site from an agent

Every page here ships in two forms — one for people, one for machines: a skill, RSS feeds, JSON and CSV datasets, and a Markdown twin of every detail page. All of it is anonymous, read-only, and needs no API key or sign-up.

Four ways in

Roughly ordered by how much work each one saves you:

  • Skill — a Claude-style skill file that tells your agent when to consult this site and how to cite it.
  • RSS — one item per new record, so a watcher can poll instead of crawling.
  • JSON / CSV — debates, policies, tracker, videos and one cross-domain updates feed as plain files, each carrying its license and links back.
  • Markdown twins — every debate, policy and video page with the layout stripped out.

License: sgai-authored content (summaries, translations, analysis) is CC BY 4.0 — attribute and link to sgai.md. Verbatim source text (Hansard, speeches, transcripts, policy documents) remains © its original rights holders and is reproduced for reference only. Terms: DATA-LICENSE

Skill (Claude Code)

Install it into Claude Code:

mkdir -p ~/.claude/skills/sgai
curl -L https://sgai.md/skill/SKILL.md -o ~/.claude/skills/sgai/SKILL.md
curl -L https://sgai.md/skill/url-map.json -o ~/.claude/skills/sgai/url-map.json

Restart the session and run /skills — sgai should be in the list. The skill ships no data of its own: it tells the agent which page to fetch, and requires every answer to link back to the specific sgai.md page it used.

Then ask, for example:

  • “Which Singapore ministers have spoken on AI safety in Parliament?”
  • “What AI grants can a foreign founder apply for in Singapore?”
  • “How does Singapore’s AI governance compare with the EU AI Act?”

RSS

updates.rss.xml carries one item per new record — debates, policies, videos, startups. rss.xml carries longform posts only.

https://sgai.md/updates.rss.xml
https://sgai.md/rss.xml

Both have a per-locale mirror under the language prefix — /zh/, /ja/, /ko/, /zh-tw/. For example https://sgai.md/zh/updates.rss.xml.

JSON and CSV

File What is in it
/data/index.json The dataset catalogue: every file with its URL, current row count and a one-line description. Start here.
/data/records.json Every dated record on the site, newest first, all domains merged — the one to poll for what is new.
/data/debates.json Every parliamentary debate record: id, date, type, Chinese and English title and summary, topics, speakers, source URL.
/data/policies.json Every policy document: id, date, issuing body, Chinese and English title and summary, source and PDF links.
/data/tracker.json The six tracker dimensions with their headline figures, ranking anchors and metric rows.
/data/videos.json Every catalogued talk: speaker, affiliation, topic, YouTube URL and four-language summaries. Transcripts are not included.
/data/debates.csv The debate dataset as CSV, for spreadsheets and notebooks.

Every JSON file except the CSV ships the same envelope: schemaVersion pins the shape, license travels with the data, and the rows live under count and items. Each row also carries links.sgai — its page URL in all five languages, which is what you should cite. The full contract is at openapi.json.

The newest record in the feed:

curl -s https://sgai.md/data/records.json | jq '.items[0]'

Markdown twins

Every debate, policy and video detail page has a Markdown version: take the same path, drop the trailing slash, add .md. Same content, no navigation and no styling — cheaper to read and cleaner to quote.

https://sgai.md/debates/written-answer-23654.md
https://sgai.md/ja/policies/nais-update-2026.md

llms.txt

llms.txt is a short site map written for models; llms-full.txt lists every page with a one-line description. Fetch the full one first when you do not yet know where to look.

https://sgai.md/llms.txt
https://sgai.md/llms-full.txt

Ground rules

  • Attribute, and link to the specific page you used — not to the homepage.
  • Verify numbers, policy text and quotations against the original sourceUrl on the page before relying on them. Summaries and translations here are produced by people and models together; the original document always wins.
  • Respect robots.txt. AI retrieval bots are allowed everywhere on this site, deliberately.
  • Do not mirror the site. Fetch the pages you need when you need them — the data changes weekly.

About this site