Display learnings for selection:
ALWAYS use a markdown table — never use section breaks, horizontal rules, or prose paragraphs to list learnings.
Identified learnings from this session:
| # | Learning | Type | Target File | Utility |
|---|----------|------|-------------|---------|
| 1 | LGTM verification process | Skill | commands/address-pr-review/SKILL.md | High - novel project pattern |
| 2 | Co-authorship in PR replies | Guideline | guidelines/git-workflow.md | Low - already documented |
| 3 | SessionEnd hook configuration | Learning | learnings/ci-cd.md | High - useful reference |
Target files are relative to ~/.claude/.
Utility ratings (self-assessment of value to Claude):
- High - Novel pattern I wouldn't know without documenting
- Medium - Useful reminder, but I could rediscover if needed
- Low - Standard knowledge or already documented (shown for transparency)
Auto-save High and Medium-utility learnings. High and Medium learnings are automatically included — they represent patterns worth preserving without asking.
Prompt for Low-utility learnings. Use AskUserQuestion with multi-select to let the user choose which Low learnings to also capture (if any exist). If no Low learnings exist, skip the prompt entirely and proceed to writing. Include the utility rating in each option's description field (e.g., "Utility: Low — standard knowledge or already documented").
After the table, clearly state which learnings will be auto-saved:
Auto-saving N High/Medium-utility learning(s). Select any Low learnings to also include:
Combine the auto-saved and user-selected items into SELECTED_LEARNINGS.
Do NOT proceed until user selects (or all are auto-saved). If no learnings selected and none are High/Medium, inform user and exit.
