| name | auto-learn-agent |
| description | Accelerate reading-list digestion for quanta-learn. Auto-import learning materials, match tool-list and solved-list using reading metrics, convert actionable items to problems, and maintain related cross-references. |
Quanta Learn Agent
Primary goal: digest the reading-list backlog, not only answer one-off questions.
Default workflow (reading digestion)
- List
<CATALOG_READING> items with status: inbox or active.
- Match
<CATALOG_TOOL> then <CATALOG_SOLVED> using tags, category, title, topics.
- If sufficient → set
related, status: done, optional summary.
- If needs work and category is algorithm/debug/system-design → ensure problem exists (
reading_to_problem.py), solve, then update solved/tool and related.
Secondary workflow (new question)
Search order: tool-list → solved-list → reading-list → problem-list.
Ingest pipeline
export CHROME_USER_DATA_DIR="<your-browser-profile-dir>"
python3 scripts/import_chrome_sources.py
python3 scripts/classify_reading_items.py
python3 scripts/reading_to_problem.py
python3 scripts/sync_catalog_from_legacy.py
Never write to the browser profile. Do not commit local catalog YAML or reading snapshots.
After solving
- Update reading
related.solved / related.tools
- Add or refresh solved-list; propose tool-list if reusable
