| name | peter-zhou-overview |
| description | Compute Peter Zhou learning overview reports. Use for total wrong counts, subject breakdown, mastery progress, repeat-correction markers, weak knowledge tags, correction candidates, and recent source/correction activity. |
Overview
Use this subskill when the user asks for progress, mastery, weak points, or what to practice next.
Core contract:
- Compute reports on demand from canonical JSON stores with
scripts/knowledge_overview.py overview.
- For Agent-chat status, progress, general Peter Zhou startup, or “what next” requests, render the compact dashboard with
scripts/chat_dashboard.py --data-dir <dir> --top-n 3. Send its Markdown directly, preserving emoji, action numbering, and links to existing local assets.
- When the Dashboard offers
⏱ 今日 10 分钟复习, route it to the resumable agent_chat_review workflow rather than generating a paper.
- The dashboard discovers deep KnowledgeModule courseware from validated runtime metadata. It must ignore unvalidated metadata, missing HTML, and references that resolve outside
data/; do not add these modules to the lightweight knowledge-pages/index.json.
- Check
scripts/learning_profile.py show --data-dir <dir> --json when the user asks for personalized priorities or study advice. If the profile is missing, overview can still run, but personalized recommendations should ask the first-use interview first.
- Count only
is_wrong=true records for final wrong-question statistics.
- Derive mastery from correction counters; do not maintain aggregate tables.
- Sort weak tags by unmastered count, then low tag mastery, then total wrong count.
