LLM-maintained context¶
| Field | Value |
|---|---|
| Status | Active |
| Type | Practice / workflow |
Description¶
Using an LLM (especially inside Cursor) as the tireless maintainer of a personal wiki: summarizing, linking projects, updating statuses, and recording decisions so you rarely hand-edit wiki/ yourself. You supply raw material and direction; the model updates index.md, log.md, project pages, and ../What-Im-Working-On.
Key features¶
- Ingest:
raw/or pasted text → structured pages + log entry. - Query: Answers should flow back into the wiki when they contain durable truth.
- Explicit refresh: “Update What-I’m-Working-On” triggers a tight rewrite of the public hub.
- Lint: Stale statuses, orphans, missing cross-refs.
Tech stack¶
- Cursor + chat; optional rules pointing agents at
SCHEMA.md.
Raw sources¶
- Pattern spec:
SCHEMA.md(repo root).
Related¶
- personal-projects-wiki — repository that implements this pattern.
- Authoritative instructions:
SCHEMA.mdat the repository root (add a vault note or symlink if you want a wikilink in Obsidian).
Known issues¶
- Quality depends on consistent prompts and occasional human review of
What-Im-Working-Onbefore public publish.