| name | fetch-doenet-activity |
| description | Get the DoenetML source behind a beta.doenet.org / doenet.org activityViewer link. The pages are a client-rendered SPA, so fetching the URL (WebFetch/curl) returns an empty shell — call the content JSON API instead. Use when asked to read, extract, or reference the DoenetML of a Doenet activity link. |
Fetching DoenetML from a Doenet activity link
Use this skill when you need the DoenetML source behind a Doenet activity URL such as https://beta.doenet.org/activityViewer/<contentId>, where <contentId> is the trailing path segment (e.g. 5pDL36qo1gJsogpTjQxVnM).
Why the obvious approach fails
The activity pages are a client-rendered React single-page app served from S3/CloudFront. Fetching the page HTML (WebFetch, curl, etc.) returns only the empty app shell — the title "Doenet" and no activity content. The DoenetML is loaded at runtime from a JSON API, so you must call that API directly.
Fast path (verified on beta.doenet.org)
Call the content API with the id from the URL and read .activity.doenetML:
id=5pDL36qo1gJsogpTjQxVnM
curl -s "https://beta.doenet.org/api/activityEditView/getActivityViewerData/$id" \
| python3 -c "import json,sys; a=json.load(sys.stdin)['activity']; print(a['doenetML'])"
The same JSON also carries .activity.name (the human title). The endpoint is public — no auth needed. Related routes in the same family, seen in the site bundle: …/api/activityEditView/getContentSource/<id> and …/api/activityEditView/getPublicContent/<viewId>.
If the endpoint 404s or the host differs (self-healing)
This API belongs to the Doenet website (a separate product from this rendering-engine repo), and beta is a moving target: the route can change, and production doenet.org may differ. Re-derive it from the site's JS bundle rather than guessing:
curl -s https://beta.doenet.org/activityViewer/<id> | grep -oE '/assets/index-[^"]+\.js'
curl -s https://beta.doenet.org/assets/index-XXXXXXXX.js \
| grep -oE '`?/api/[a-zA-Z0-9/_.${}-]+' | sort -u | grep -iE 'activity|content|view'
Find a get…ViewerData / getContentSource / getPublicContent route, call it with the id, and inspect the JSON for the field holding the source (it was activity.doenetML on beta).
Caveats
- This is the website/beta API, not part of this repo — treat it as external and subject to change. If the literal URL above fails, trust the self-healing grep over the cached route.
- Fetches public content only. Private or unshared activities need authentication this skill doesn't cover.
