OPEN RESEARCH · HTTP / AGENT SKILL
Bring research into your workflow
Pages, scripts and AI assistants share the published research. Content retains sources and dates.
Start with one query
- For scripts, copy the curl example below. For an AI assistant, follow the Skill installation instructions. Both access the same public content.
- Search titles and summaries first, then fetch the articles or company snapshots you need. Avoid downloading the whole site.
- Include the original article link, publication date and key sources so readers can check the context.
Request convention: cache first; one request at a time, at most 6 per minute and at least 10 seconds apart.
Stop on HTTP 429 or access denial and inspect the response and Retry-After. Do not retry automatically, rotate IPs or poll continuously. This is a client convention; there is no per-client quota counter currently.
01 / HTTP API
Query published research
| Endpoint | Purpose |
|---|---|
GET /api/v1/status | Content counts, availability and snapshot dates |
GET /api/v1/articles | Search articles; lang, q, tag, page, page_size |
GET /api/v1/articles/{slug} | Article metadata and sanitized HTML content; lang |
GET /api/v1/radar | Market pulse, sources and automated analysis |
curl --fail 'https://growthyoung.com/api/v1/articles?lang=en&q=backtest&page_size=5'
lang: zh-CN (default) or en; page starts at 1; page_size defaults to 20, maximum 50. q accepts up to 100 characters. tag uses stable IDs listed in OpenAPI.
Responses, caching and errors
Success returns ok, data and meta; failures return ok: false, error and detail. Invalid, repeated or unknown parameters return 422. Unpublished articles return 404, unsupported methods 405, and missing or invalid market pulse data 503, distinct from a valid empty event set.
Lists sort by publication date descending, then slug. New content may move pagination; deduplicate by slug. Data and specifications support 60-second caching and ETag; status is not cached.
02 / AGENT SKILL
Let your AI read the research
The bundle contains SKILL.md and an API reference. It guides the assistant to search, read the full article and preserve dates and evidence limits. It uses HTTP directly, without MCP.
Install the strategy-research skill in the current AI client to query published research and market pulse. https://growthyoung.com/skills/strategy-research.zip Identify the client and inspect the archive before installing in its supported skill directory. Verify strategy-research/SKILL.md and references/api.md; preserve local customizations and other skills. Read the skill and verify an article query with the X-Research-Client: strategy-research-skill header, requesting serially at least 10 seconds apart. Preserve article links and dates in answers. Report the path, actual verification results and any client reload needed.
- Find articles on AI governance, with sources and revision dates.
- Read the backtesting guide and check the execution assumptions.
- Read the Rocket Lab pulse and distinguish reports from automated analysis.
03 / SOURCES AND UPDATES
Keep dates with your citations
published_at and updated_at are publication and revision dates; market pulse generated_at is collection time, and analysis.generated_at is analysis time. served_at is only request time.
analysis.status=ok means evidence-set alignment. Disclose old dates, stale, waiting, invalid and missing data accurately. The temperature is not observed buying, a calibrated forecast or a trading instruction.
04 / USAGE AND PRIVACY
API calls and page views are counted separately
Since October 15, 2026, requests reaching the content service are aggregated by UTC+8 calendar day, route family, HTTP method and response status.
| Channel | How it is identified |
|---|---|
| API | Ordinary API requests, including requests without a Skill marker. |
| Skill | When using the Skill, send X-Research-Client: strategy-research-skill. This is self-declared attribution, not an API key or authentication. |
| Skill files | Visits to the ZIP, SKILL.md and reference files; a download does not prove installation or execution. |
The three channels are mutually exclusive and are not browser PV. Totals include failed responses, HEAD and OPTIONS; cache hits that do not reach the content service are excluded. Channel history before collection began cannot be reconstructed.
These daily aggregates do not store IP addresses, User-Agent, visitor IDs, query content or raw header values. Clients without custom-header support can still use the API; their requests count in the API channel.
05 / SUPPORT AND COLLABORATION
Help useful research keep growing
This site is independently maintained. Preserve original links when citing research, share useful articles and report missing evidence or errors. Readers choose whether to visit the original article or support the site.
Web pages may display ads. API and Skill access does not require ad views or clicks; clients must not load, refresh or click ads automatically in the background.
For research collaboration, content licensing or sponsorship enquiries, get in touch by email. This page currently offers no paid plans or subscription checkout.