brightspace-notes
/brightspace-notes
Read and navigate a Brightspace (D2L) course, capture its lecture videos (subtitles + one frame per slide) and documents, and write revision notes as .md files per topic with timestamp-linked slide references. Use when the user asks to capture, summarise, or make revision notes from Brightspace/e-learning material or lecture videos.
- Author
- arz67mangos ↗
- Repository
- brightspace-notes ↗
- Licence
- MIT
- Last repository change
- 30.10.2026
Assessment and security
- Files in this package
- 1
- Scripts
- 0
- HTML comments
- 0
- URLs in SKILL.md
- 0
No signals detected. Repository changed on 30.10.2026, without references to retired models or APIs.
MIT, open: use and modify while retaining author attribution.
What it does and when to use it
Read and navigate a Brightspace (D2L) course, capture its lecture videos (subtitles + one frame per slide) and documents, and write revision notes as .md files per topic with timestamp-linked slide references. Use when the user asks to capture, summarise, or make revision notes from Brightspace/e-learning material or lecture videos.
Original author description
Read and navigate a Brightspace (D2L) course, capture its lecture videos (subtitles + one frame per slide) and documents, and write revision notes as .md files per topic with timestamp-linked slide references. Use when the user asks to capture, summarise, or make revision notes from Brightspace/e-learning material or lecture videos.
How it works
- 0. Setup
- 1. Scope first: never crawl everything
- 2. Videos
- 3. Documents (file-activity)
- 3b. Topic types and how to handle each
- 4. Write the note
- 4b. Verify coverage (do this after every capture session)
- 5. Revision outputs (on request)
- 5b. Course book with blanks to fill in (printed notes with and empty table cells)
- 6. After every capture session: refresh the agent context
SKILL.md preview
---
name: brightspace-notes
description: Read and navigate a Brightspace (D2L) course, capture its lecture videos (subtitles + one frame per slide) and documents, and write revision notes as .md files per topic with timestamp-linked slide references. Use when the user asks to capture, summarise, or make revision notes from Brightspace/e-learning material or lecture videos.
---
# Brightspace -> revision notes
Repo root = the folder containing `scripts/` and `notes/`. Run everything from it. Notes: `notes/<course>/<module>/<topic>.md`; each topic's captured media: `notes/<course>/<module>/assets/<slug>/` (all gitignored).
Related skills: `brightspace-tutorial` (answer tutorial sheets from the notes), `brightspace-quiz` (quiz the user from the notes).
## 0. Setup
- First time on this machine: follow the `brightspace-setup` skill (checks what's installed, asks before installing anything). Something broken later: `brightspace-check`. Both use `python scripts/doctor.py`, which prints exactly what's missing and the fix.
- `python scripts/bs.py login <base-url>`: opens Chromium, the user signs in (SSO/MFA), the script saves the session and exits by itself. Run it in the background and tell the user to sign in. Re-run when the session expires (symptom: redirect to a login page).
**Browser window:** by default the scripts open a visible Chromium window so the user can watch. The first time you run a browser script in a session, tell the user once: "A browser window will open so you can watch; if you'd rather it ran hidden, say 'switch to headless mode'." When they ask to hide it run `python scripts/bs.py mode headless`; to show it again run `python scripts/bs.py mode show`. Don't touch the window while a script runs. The login window is always visible.
## 1. Scope first: never crawl everything
Ask for the base URL and course (or topic). A pasted course URL contains the orgUnitId (`/d2l/home/<id>`). Run `python scripts/bs.py toc <base> <ou>` (one request, writes `notes/<ou>/toc.md` with every module/topic, its activity ID, and kind), show the modules, and ask which ones they want (AskUserQuestion, multi-select). Process only those.
- **Only capture what was asked for.** Skip quizzes, surveys, assignment dropboxes, group lists and external links unless the user asks (quizzes: they may want help with participation/pre-class quizzes or self-tests; see `brightspace-quiz`). Never submit, start a timed attempt, or click anything that records a result on the user's behalf.
- Reuse an existing `toc.md` instead of re-fetching. `link-activity` = usually a video (or a quiz: check the title), `file-activity` = a document.
- Interactive alternative or when scripts break: Claude in Chrome, if the user has it connected.
## 2. Videos
1. `python scripts/bs.py video <base> <ou> <activityId> notes/<course>/<module>/assets/<slug>`: finds the player through the adapters in `scripts/hosts.py` (YuJa is tested; YouTube/Vimeo/Panopto/Kaltura embeds and generic sniffing are untested), writes the caption file and `source.txt` (signed stream URL, expires in about an hour).
2. `python scripts/video.py "$(cat <dir>/source.txt)" <dir>` then delete `source.txt`. If step 1 printed `"host": "embed"`, add `--cookies cookies.txt` (written by step 1, gitignored; delete it when done). Produces:
- `slides.md`: one section per distinct slide: frame link + the narration said while it was on screen. **Start here.**
- `sheets/sheetNN.jpg`: contact sheets of 4 slides each, labelled `#n mm:ss`. **Read these to see every slide cheaply**, then open single `frames/` jpgs for small print, graphs and tables.
- `transcript.md`, `frames.txt` (file, start, end, sheet cell).
- Slides are not missed: the video is sampled every 2 s and a frame kept whenever the picture changes (captions at the bottom are ignored). If bullets appearing one at a time are being missed, rerun with `--sim 0.02`; if too many near-duplicates, `--sim 0.06`. Topics with 2+ videos: one asset folder each.
3. Subtitles: soft subs are used when they exist. If a video has none, `video.py` transcribes the audio with faster-whisper (`--whisper-model small.en` is better for coursework).
4. **Fix transcript errors**: captions are ASR. Cross-check every term against the slide text in the frames (and the lecture-notes document if captured). Correct the term in the note and list the correction under Gaps (e.g. "demand" misheard as "D-man"). Treat lines that contradict the slide as ASR errors.
5. Other hosts: if `bs.py video` finds nothing, see CONTRIBUTING.md (add an adapter) or in Chrome use `read_network_requests` to find the `.m3u8/.mp4/.vtt` URLs and pass them to `video.py`.
## 3. Documents (`file-activity`)
`python scripts/bs.py doc <base> <ou> <activityId> work/<slug>` downloads the file; then `python scripts/docs.py <file> notes/<course>/<module>/assets/<slug>` converts pdf/docx/pptx/zip to markdown (+ page images for scanned/diagram pages, or all pages with `--images`). Read the `.md`; view page images for diagrams. Lecture-notes documents are the authority for terminology and model diagrams.
## 3b. Topic types and how to handle each
**Slide decks (PDF/PowerPoint shown in Brightspace's viewer): never screenshot pages.** Download the file with `bs.py doc`, then `docs.py <file> <dir> --images` (text + page images; for `.pptx` without LibreOffice you get text only: say so, offer to install LibreOffice or to use the PDF version if one exists). Work from the text and look at page images only for diagrams.
**YouTube/Vimeo links** (a page that just lists video links): get the page's links (`docs.py` on the downloaded page, or `bs.py page`), then run `video.py <youtube-url> <dir>` per video: same transcript + slide-frame method (yt-dlp fetches the captions).
**Interactive games / activities** (drag-and-drop, simulations): the learning is in doing, so don't try to "scrape" a game. Read its instructions/text, note the concept it teaches, and ask the user how they'd like it handled (options: they play it and tell you what they saw; you open it visibly and screenshot the states; you only summarise the intro text). Never press submit/grade.
**Articulate/Rise SCORM courses** (lessons "Video 1..N" + a Review Quiz, each lesson holding a YuJa video and short self-test questions; `bs.py video` finds no stream): run `python scripts/bs.py course <base> <ou> <aid> <dir>` (clicks each lesson, writes `lessons.json` with each lesson's text, the course-book sections/pages it covers, and its YuJa URL; never touches the Review Quiz), then `python scripts/bs.py yuja <base> <url> <dir>` + `video.py` per URL. Lessons that link an external video (library streaming sites) have no `video` in lessons.json: list them for the user.
**Other multi-page interactive courses** (navigate page by page with videos, slide notes and short self-test questions): **new situation: stop and ask first.** Describe what you see and give 2-3 options with your recommendation, for example: (a) step through each page with the browser and record text, embedded videos and self-test questions; (b) check whether the course package can be downloaded; (c) the user pages through while you read each page as they go.
**Any page type you don't recognise:** don't improvise. Tell the user what it is, give 2-3 options with your pick, and wait for the answer.
## 4. Write the note
One file per topic. Front matter lists the videos so citations can be turned into frame links:
```
---
course: <name>
module: <name>
topic: <title>
source: Brightspace activity <id>
captured: <YYYY-MM-DD>
video: P1 | Part 1 Law of Demand | assets/part1
---
# <Topic>
## Key ideas (bullets; **bold** each key term: they feed GLOSSARY.md)
## Details (formulas, worked examples, diagrams as text; cite the video time: [02:10] or [02:10-02:40])
## Likely exam/test points
## Self-check questions
1. Question? *Short answer*
## Gaps (ASR corrections, unreadable frames, anything not captured)
```
- Cite video times as `[mm:ss]` / `[mm:ss-mm:ss]`; with 2+ videos in the note write `[P2 02:10]`. Then run `python scripts/linkify.py <note.md>`: each cite becomes a link to the slide on screen at that time (`[Part 1 @02:10](assets/part1/frames/f003_...jpg)`), so the user can jump back to the slide.
- Self-check questions MUST be `N. Question? *answer*` (answer in italics, last): that is what `anki.py` exports.
- Never invent content that isn't in the transcript/frames/documents. Unknowns go under Gaps.
- Update `notes/<course>/INDEX.md` (one line per note).
## 4b. Verify coverage (do this after every capture session)
`python scripts/verify.py <ou> notes/<course>` writes `notes/<ou>/COVERAGE.md`: every topic in the outline with its status (captured / FAILED / not captured, quiz-like ones labelled), and per capture the evidence: slide count, video length vs transcript range (flags transcripts that stop early or have long silent gaps), missing frames, documents downloaded but not converted, PDFs whose page count doesn't match, plus broken links, unlinked citations and self-check questions without answers in the notes. Fix whatever it flags, then tell the user the headline ("12 of 16 topics captured; 3 not requested; 1 failed: ...") and anything it can't see (other Brightspace tools; items not yet released). Every `bs.py video/doc` attempt is logged to `notes/<ou>/capture-log.jsonl` for this.
## 5. Revision outputs (on request)
- `python scripts/rollup.py notes/<course>` -> `EXAM_FOCUS.md` (all exam points), `GLOSSARY.md` (all bold terms), per-module `CHEATSHEET.md`. Offer to tighten a cheat sheet into a one-page summary.
- `python scripts/anki.py notes/<course>` -> `notes/<course>/anki-cards.tsv`, import in Anki (File > Import).
- Quizzes, self-tests and tutorial help: skills `brightspace-quiz` and `brightspace-tutorial`.
## 5b. Course book with blanks to fill in (printed notes with ____ and empty table cells)
1. `python scripts/docs.py <book.pdf> notes/<course>/coursebook/source` for the text; render pages with images (`pymupdf` `get_pixmap`) to see tables, graphs and blanks that text extraction loses.
2. The videos usually say which book section and pages each lesson covers (the SCORM lesson text, e.g. "Course book Chapter 1: section 4.3, pages 12-15"): fill a page's blanks from that lesson's transcript and slides, then from the book's own printed text (mark those *(book text)*), never from outside knowledge. Recompute numbers from the course data files where the book prints partial results, and check they match the printed ones.
3. Write `notes/<course>/coursebook/chapter-N-filled.md` per chapter: page by page, answers as `**[[answer]]**`. Then `python scripts/mdtodocx.py <file.md>` makes the Word copy (answers in blue bold).
4. **Quiz check**: Brightspace "hands-on activity" / "practical exercise" items are often timed, graded quizzes (open the item's summary page only, never start it). If a blank matches a quiz question (same question in the book), do not fill it; say it is withheld and ask.
5. Leave chapters with no released material unfilled and say so.
## 6. After every capture session: refresh the agent context
1. `python scripts/verify.py <ou> notes/<course>` (coverage), `python scripts/rollup.py notes/<course>` (exam focus, glossary, cheat sheets).
2. `python scripts/master.py notes/<course> --ou <ou>` writes `MASTER.md`: the file any agent reads first when revising (rules: answer only from the notes, cite them, say "not covered" instead of inventing; a topic map saying which note, key terms and source files hold each topic; a file tree).
3. Write or update `notes/<course>/TIMELINE.md` from the captured schedule, assessment and reminder pages (not the lecture notes): a table of **When | What | Type (lesson F2F/SDL, quiz, CA, presentation, submission, exam, other) | Weight | Source (link to the captured file) | What to prepare with (links to notes)**, plus a "Not stated / to confirm" list. Use only dates and weightings stated in those documents; if only a week number is given, show it as given (and the term start date if the schedule has it). Never guess a date. Mention the source file for every row. Reminder pages are read for dates only; never open or answer the quizzes they point to unless the user asks.
4. Tell the user what changed and where `MASTER.md` and `TIMELINE.md` are.
## Rules
- Only courses the user is enrolled in. Don't redistribute course material; `notes/`, `.profile/`, `cookies.*` and `work/` are personal and gitignored. Never print or commit credentials; don't bypass the pre-commit hook.
- Ask before downloading in bulk; work module by module. Opening a topic marks it viewed in Brightspace.
Versions
The database has no other skills named brightspace-notes.
Database snapshot: 13.11.2026 · package version d176955d0953 (SHA-256 digest of the entire package)
Downloads in the last 30 days: 0 · ranking
Report a problem with this skill: harmful content, incorrect descriptions or an author’s removal request.
Similar skills
- 86brightspace-check
Check that the brightspace-notes setup is healthy (Python, ffmpeg, packages, browser, privacy hook, Brightspace login) and fix what's missing, explained in plain language. Use when the user says "check my setup", "is everything ok", "it's not working", or runs… - 90brightspace-setup
First-time setup of brightspace-notes on a new machine. Detects what is already installed (OS, Python, Python packages, Playwright Chromium, ffmpeg, Deno, LibreOffice, privacy hook), explains it in plain language, asks before downloading or installing each item, installs only… - 100lecture-notes
Turn a video lecture (YouTube or any yt-dlp URL, or a local video file) into one faithful Obsidian Lecture note where every idea cites a timestamp or Key frame, checked by `lecture-notes verify`; or into its full text with the slides, code and formulas in place. Use when the… - 100lecture-to-notes
Converts supplied slides, transcripts, readings, or messy notes into structured revision notes with source-vs-inference labels, a verify list, and self-test questions. Use when the learner pastes lecture material, asks for a notes generator, or wants clean markdown from slides or docs. - 100revision-notes-generator
Generate concise, self-contained revision notes from PDF or Markdown study materials. Use whenever asked to create revision notes, study notes, topic summaries, or condensed notes from educational PDFs, Markdown files, textbooks, or course materials (e.g. Save My Exams, PMT,… - 100d2l-discussion-responder
Review D2L/Brightspace discussion threads as an instructor, identify unanswered questions, verify course facts, and draft concise replies in the instructor's voice. Use when asked to respond to student discussions or catch up on an LMS forum; post only within the user's explicit…