update-learning-knowledge

name	update-learning-knowledge
description	Search trusted sources, organize new and changed learning material into a source-traceable bidirectional knowledge base, and refresh a seven-day plan with exactly two learning cards per day. Use when maintaining the AI learning knowledge base, preparing daily knowledge updates, generating or refreshing cards, or running the scheduled daily update workflow.

Update Learning Knowledge

Maintain the learning system through one auditable pipeline: search, verify, organize, link, generate cards, and validate.

Daily workflow

1. Read references/source-policy.md before searching.
2. Search the web for material published or materially updated since the last run. Prefer primary sources and official documentation. Save normalized results using references/search-contract.md.
3. Reject entries without a direct source URL, publication date, or clear factual change. Mark uncertain claims for human review; do not insert them into the formal knowledge base.
4. Run the updater from the project root:

python3 skills/update-learning-knowledge/scripts/update_learning.py \
  --source knowledge-base/sources \
  --wiki-output knowledge-base \
  --learning-data data/learning-data.json \
  --search-results /absolute/path/to/daily-search-results.json \
  --config skills/update-learning-knowledge/assets/update-config.example.json \
  --validate

5. Review the printed summary. Confirm that the site data contains seven days, two cards per day, valid source references, a daily update preview, and no duplicate card IDs.
6. Report added, corrected, linked, and rejected items separately. Include direct source URLs for every accepted web update.

Card rules

• Generate the complete seven-day queue in one run: exactly 14 cards, two per day.
• Each card must teach a useful unit of knowledge; it may be a concept, method, process, principle, or tool.
• Keep each card traceable with at least one sourceId.
• Pair complementary cards each day: understanding + application, or principle + method.
• A daily run refreshes today's two cards only when trusted source changes justify it. Preserve learner progress and stable IDs whenever possible.
• Never copy a paper abstract as a card. Rewrite it as a question, concise explanation, concrete example, correction hint, and action.

Knowledge organization rules

• Preserve source files. Generate knowledge pages and reciprocal links in the wiki output.
• Treat corrections as corrections, not new facts. Record what changed and why.
• Keep generated blocks idempotent; retain manual notes outside generated markers.
• Do not expose this Skill, its schedule, or internal maintenance prompts in the website navigation.

Scheduled execution

Use the daily task prompt and operating rules in references/schedule.md. The task should run at 09:00 local time and write a concise run summary after validation.