Turn instructions and homework sheets into an editable task draft
AI extracts, separates and schedules; you decide. Edit subjects, interests, dates, duration, checklists and completion modes before anything is added to the real task plan.
- Skill ID: ai-task-planner
- Reasoning required; vision optional for images
- Minimum data scopes: tasks and files
30 minutes of math next week, vocabulary every other day, and less work on weekends.
Understand the sources first, then produce a draft you can reject or change
This Skill does not turn one sentence into a final calendar. The client reads explicitly selected sources, the model creates a structured schedule, and the UI keeps a complete edit and confirmation step.
Text and multiple file types as input
Use natural language, images, PDFs, TXT, Markdown, CSV, Word or Excel. PDFs are parsed as text first, with scanned pages sent through vision when needed.
Split by real learning semantics
Identify subjects, interests, dates, deadlines and teacher requirements. Group same-date subject work into checklists and separate one-off tasks from habits.
Keep the schedule fully editable
Change title, subject, interest, date, time, duration, instructions, checklist and completion mode, or continue the conversation to revise the whole draft.
Four reviewable steps between a homework sheet and local tasks
AI should reduce organization time without removing judgment. Source handling, extraction, scheduling and the final write have explicit boundaries.
- 1
Describe the request and choose sources
Ask for a schedule such as “30 minutes of math each day, less on weekends,” then attach homework sheets, notices or planning files.
- 2
Extract subjects, dates and requirements
The client parses text. Vision faithfully extracts image or scanned-PDF content without making scheduling decisions during OCR.
- 3
Generate and inspect the draft
Schedule around deadlines, subjects and daily load, with warnings when work falls after a deadline or exceeds the suggested daily duration.
- 4
Edit and confirm the final write
Adjust every item in list or calendar preview, remove weak suggestions, and create real tasks only after confirmation.
Structured tasks instead of disposable advice
Explicit classification
Keep school subjects separate from interest practice and match existing interest records where possible.
An executable schedule
Every item carries a date, time, estimated minutes, priority, recurrence and checklist rather than vague advice.
A traceable source
Confirmed tasks preserve the AI_PLAN source and draft identifier so they remain distinguishable from manually created work.
The Skill reads authorized scopes and proposes; it cannot bypass confirmation
The backend registers AI task planning as an independent Skill with tasks and files scopes. Reasoning is required and vision runs only for image sources. Platform or configured local models still return a draft first.
AI task planning ultimately feeds the task planner. Assignment, receipts, reviews and recurring check-ins remain core product responsibilities.
→Test it with a real homework sheet, not a polished claim
Upload one schedule, inspect how AI separates dates, subjects and checklists, edit weak suggestions, and only then decide whether to add the tasks.
Download the client