Use teach-back for spaced review with an agent
Explain a topic before seeing the answer, work through one gap at a time and optionally save reviewed cards for later study sessions.
Use teach-back for spaced review with an agent to begin with what you can explain unaided. The study-loop skill picks one gap in your account and asks a focused question, rather than delivering a full lesson immediately. Repeated difficulty leads to graduated help, with useful points retained as possible cards.
You can edit or discard the proposed cards before agreeing to save them. Later reviews show a question before its answer, then use your own right, almost or wrong assessment to adjust a five-box schedule. Due dates organise future sessions; they are not automatic notifications or evidence of lasting mastery.
What it does
- Explain a topic in your own words.
- Work on one missing step or assumption.
- Review a small set of proposed cards.
- Approve local storage after checking the wording.
- Answer due cards before revealing their backs.
- Adjust review dates through self-assessment.
Install and use
- Use the complete package in a file- and command-capable agent. Python 3.8 or newer is needed for storage and scheduling; without it, only the conversation continues. Avoid simultaneous saves for one topic and choose clearly distinct topic names.
- For example, offer your explanation of an authorised water-cycle paragraph and ask the agent not to reveal the key point before you answer.
- Check proposed cards and the reported storage location, then approve only those you want. A failed save must remain a failed save in the report.
- Return later to request due cards. Answer, inspect the reference answer and supply your assessment. Optional Context7 checks for library/API topics send generic questions, not private learner text.
FAQ
Source
Made by lpalvess. Original skill: study-loop, published on GitHub under the MIT licence.
View on GitHub