brightspace-triage

name	brightspace-triage
description	Query Brightspace (D2L) for grades, due dates, and assignments, then compute and present a prioritized study plan. Use when the user asks what to work on, what's due, how they're doing in a class, or asks for a prioritized to-do list. Also use for formatting grade summaries and weekly planning.

Brightspace Triage

Turn raw Brightspace data into a ranked, explainable study plan.

When to load

Trigger when the user:

• Asks "what should I work on?" / "what's most important?"
• Asks for grades, deadlines, or upcoming assignments
• Wants a weekly or daily study plan
• Mentions Brightspace, D2L, or a specific course
• Invokes /grades, /week, or /prioritize

Tools you have

The brightspace-mcp-server MCP exposes (names are approximate; discover real ones at runtime):

• get_my_courses — enrolled courses
• get_my_grades — grades per course (numeric + letter)
• get_upcoming_due_dates — deadlines across all courses
• get_assignments — assignment list for a course (with points, due, submission status)
• get_syllabus — course syllabus (used to extract grade weights)
• get_announcements, get_discussions, get_course_content, download_file

Prefer get_upcoming_due_dates + get_my_grades + get_assignments for triage. Batch calls — don't serialize one per course if a single call returns everything.

The priority algorithm

Do NOT compute priority scores yourself. Always delegate to the bundled script:

python3 scripts/prioritize.py < input.json

Why a script: the algorithm weighs urgency × grade impact × risk ÷ effort. Doing this inline produces inconsistent results; the script produces deterministic, auditable output the user can tune.

Input shape (input.json):

{
  "target_grade": 90,
  "now": "2026-04-22T05:30:00Z",
  "config_path": "/home/<user>/.config/brightspace-triage/config.json",
  "courses": [
    {
      "name": "CS 23200 Intro to C and UNIX",
      "current_grade": 87.3
    }
  ],
  "assignments": [
    {
      "course": "CS 23200 Intro to C and UNIX",
      "name": "HW5",
      "category": "homework",
      "due": "2026-04-25T23:59:00Z",
      "points_possible": 50,
      "points_earned": null,
      "submitted": false,
      "effort_hours": null
    }
  ]
}

Keys you don't know → emit null and let the script fall back to defaults.

Output shape:

{
  "ranked": [
    {
      "name": "HW5",
      "course": "CS 23200",
      "priority_score": 12.34,
      "due": "2026-04-25T23:59:00Z",
      "reasoning": "due in 3.2d | 5.0% weight | 3pt below target | 3h effort"
    }
  ]
}

See references/algorithm.md for the full scoring formula and how to tune it.

User config

Optional file at ~/.config/brightspace-triage/config.json:

{
  "target_grade": 90,
  "courses": {
    "CS 23200 Intro to C and UNIX": {
      "weights": {"homework": 30, "exams": 50, "final": 20},
      "effort_defaults": {"homework": 4, "exams": 10}
    }
  },
  "global_effort_defaults": {
    "homework": 3, "exam": 8, "quiz": 1, "project": 10,
    "discussion": 0.5, "reading": 2, "final": 15
  }
}

If the file doesn't exist, the script uses built-in defaults (target_grade=90, generic effort table). Offer to help the user create/edit it when:

• They ask why a specific item is ranked where it is
• Priorities feel off to them
• They mention category weights from a syllabus

Bootstrap shortcut: the /brightspace-sync command reads every course's syllabus via MCP and proposes a merged config.json with category weights extracted. Recommend it the first time you notice the config file is missing, or when rankings feel consistently off because weights are unknown.

Formatting the output

Default to a table, not a bulleted list. Students scan tables faster.

| # | Course    | Assignment | Due            | Points | Score |
|---|-----------|------------|----------------|--------|-------|
| 1 | CS 23200  | HW5        | Fri 11:59pm    |    50  | 12.3  |

Include a short "why #1" sentence under the table citing the script's reasoning field. Show the top 5–10, not the full list, unless asked.

For a grade summary (/grades), group by course with current grade, recent graded items, and a one-line "you're N points from an A" callout.

For weekly view (/week), group by day of the week, with scores as a secondary hint.

Hard rules

• Never invent numbers. If a field is missing, show "—" and say so. Don't guess.
• Always surface assumptions — if you inferred the target grade (user didn't specify), say so.
• Don't double-count submitted work — the script filters submitted items out; if something shows up as submitted in the MCP data, trust that.
• Gradebook categories vary. If a course's categories look weird (e.g. "Participation" worth 40%), flag it to the user.