remember

<!-- tested with: claude code v2.1.122 --> When the user runs `/lore:remember`, your job is to persist what they want to keep into the `notes` table of `~/.claude/lore/lore.db`. This is the user-controlled layer of the knowledge graph -- the parts of memory that can't be inferred from JSONL alone. ## locate notes.py The CLI is shipped with the plugin. Resolve it once and reuse the path. `````bash NOTES="" for p in "${CLAUDE_PLUGIN_ROOT}/scripts/notes.py" \ "$HOME/.claude/plugins/marketplaces/anipotts/claude-code-tips/plugins/lore/scripts/notes.py" \ $(find "$HOME/.claude/plugins" -path "*/lore/scripts/notes.py" 2>/dev/null | head -1); do if [ -n "$p" ] && [ -f "$p" ]; then NOTES="$p"; break; fi done if [ -z "$NOTES" ]; then echo "could not find notes.py -- is the lore plugin installed?"; exit 1; fi echo "using: $NOTES" ````` If the resolution fails, tell the user and stop. Don't fall back to writing raw SQL -- the helper validates types and timestamps. ## intents Map the user's request to one of these intents. If the input is ambiguous (e.g. they say "remember the auth work was hard"), default to `add` with `--type lesson`. | user phrasing | intent | call | |---|---|---| | "remember <X>", "save <X>", "log decision <X>" | add | `notes.py add "<title>" [--body ...] [--type ...] [--tags ...] [--project ...] [--session ...] [--file ...]` | | "show notes", "list lessons", "what did i remember" | list | `notes.py list [--type ...] [--project ...] [--tag ...