revision-mode - SKILL.md Agent Skill

name: revision-mode description: Use when a learner needs a structured review session prioritizing high-value topics by recency, error history, and upcoming deadlines. version: 1.1.0 authors: - edu-agent-skills contributors tags: [productivity, revision, exam-prep, retention] status: stable

Purpose

Run a structured, prioritized revision session targeting topics most likely to decay, most frequently missed, or most important for an upcoming deadline. Not random review — weighted prioritization to maximize learning per minute.

Activation

Upcoming deadline (interview, exam, demo) within 1–5 sessions. Learner requests review after a learning period. weak-area-tracker shows accumulated weak areas. 3+ sessions since last review.
Skip if: learner needs new material first. Goal is project implementation → build-with-me. Only one topic to review → use check-understanding/challenge-generator directly.
Routing: pull priorities from weak-area-tracker. Use check-understanding and challenge-generator as execution vehicles. After 3 consecutive clean revision passes on a topic: mark revision-complete and de-prioritize.

Inputs

Topics covered (from learning-memory), active weak areas, upcoming deadline + type, session time budget, self-assessed confidence per topic (optional).

Priority Scoring (1–10)

Weak area severity: 0–3 (from tracker). Recency (days since last review): 0–3. Deadline relevance: 0–2. Low self-reported confidence: 0–2.
Score 7+ = must-revise. Score 4–6 = should-revise. Below 4 = defer.

Workflow

Scope — Gather topics + weak areas. Ask: "Deadline? Least confident topics?" Apply priority scoring.
Plan — Present focused list: must-revise first, then should-revise if time permits. Cap at 4–5 topics max. State approach per topic: concept check / challenge / flashcard drill.
Recap — For each topic: ask learner to summarize in 2–3 sentences. Don't re-explain unless summary reveals a gap. 2–3 min cap per recap.
Test — 1–2 exercises per topic via check-understanding or challenge-generator. Mix modes. Run 3–5 due flashcards per topic if available.
Check Exit — Clean pass = correct response with correct reasoning on first attempt. After 3 clean passes across sessions: exit rotation. Below 50% pass rate this session: escalate to misconception-detector.
Close — Summarize: which topics are revision-complete, which need another round. Update weak-area-tracker. Recommend next session if deadline still approaching.

Rules

DO: revision is active — learner produces answers, not just reads.
DO: include weak areas in every revision session until resolved.
DO: state clean-pass criteria before the session, not after.
DON'T: re-lecture topics the learner can summarize correctly — test immediately.
DON'T: pack more than 5 topics per session.
DON'T: skip the hardest topics in favor of comfortable ones — must-revise (7+) first.
DON'T: let same topics appear indefinitely — enforce exit criteria.

Output

Responses should contain: deadline + time budget, prioritized topic list (must-revise with scores and approach, should-revise if time permits), per-topic recall + test + outcome, and session summary (clean passes, still-needs-revision, exiting rotation, next session recommendation). Format naturally.

Checklist

Priority scoring applied before session plan.
Must-revise topics addressed first.
Every topic has at least one active retrieval exercise.
Weak-area-tracker updated at session close.