docs: add LLM novel writer application note

- New single-file application note at application-notes/llm-novel-writer.md demonstrating the exact 6-step pipeline (read the data if it needs → generate context → user prompt/selection → analyse change to the data → update the data → loop) with all novel data (background, configs, bibles, style rules, user preferences, current story) kept inside MemNet.
- Schema + complete initial seed (persistent reference rows + transient story rows with correct recycle values) as copy-pasteable blocks.
- 3 compact worked turns, each structured strictly as the 6 numbered steps, with warm output (including prepended LAW rows), exact memnet add/update commands, id discipline, and canon-changing updates to persistent rows.
- Mermaid diagram of the pipeline loop + persistent-vs-transient legend.
- Snapshot for full project state, pipeline-aware pitfalls, and quick-start bootstrap commands (heredoc for schema, session open, add seed, first warm).
- Updated README.md with "Application notes" section.
- Updated LLM-GUIDE.md with pointer to the example.

This is repo documentation only (single .md under application-notes/); the distributed package is unchanged.

Co-authored-by: Cursor <cursoragent@cursor.com>

Author: chouswei

LLM-GUIDE.md

@@ -263,6 +263,8 @@ memnet query warm --anchor PLR01
 - `memnet guide --loose` — short cheat sheet.
 - `memnet examples map|workflow`
 
+**Application note:** see `application-notes/llm-novel-writer.md` for a long-running creative writing example that follows the explicit 6-step pipeline (read → context → user input as data → analyse citing rows → add/update with correct recycle → loop) where background, configurations, bibles, rules and user preferences are kept as persistent rows inside the graph.
+
 **Read this file (`LLM-GUIDE.md`) at the beginning of any non-trivial task.**
 
 When the current schema or examples change, re-run `memnet examples map` and `memnet tagmap show`.

README.md

@@ -195,6 +195,12 @@ All caps have `MEMNET_MAX_*` names; see source for the full list and defaults.
 
 No JSON on the wire for LLM consumption — only the `@TAG:` lines plus a handful of control records on stderr.
 
+## Application notes
+
+Rich, document-style worked examples live under `application-notes/`. These are self-contained narratives (one `.md` per note) showing complete usage patterns with all data — including background, configurations, bibles, rules, and user preferences — kept inside MemNet.
+
+See [application-notes/llm-novel-writer.md](application-notes/llm-novel-writer.md) for the first example: an interactive LLM novel writer driven by the explicit 6-step read → context → user-input-as-data → analyse → update → loop pipeline.
+
 ## Development
 
 ```powershell