LLM Wiki Index Creation and Management

Problem

Large LLM Wikis often suffer from scattered pages, drift between index.md and actual files, and missing operation history, making it hard to trace knowledge sources and change paths. Separating index.md and log.md lets an LLM build global context with a small payload while retaining an auditable operation trail.

Core Capabilities and Key Steps

The skill centers on two core files:
- index.md: a content-oriented index that uses category grouping, one-line summaries, and metadata such as tags, status, and updated to help LLMs locate relevant pages.
- log.md: a process log that records context, timestamps, and sources for operations such as INGEST, QUERY, and EDIT, enabling debugging, auditing, and incremental learning.

Key steps include reading index.md during initialization; narrowing scope with categories or tags during task planning; consulting log.md after content retrieval to judge freshness; and updating pages, index, and log together when writing, so the file system, index, and log stay consistent. For larger collections, use layered indexing, such as root indexes, category indexes, and dynamic views like recent-updates.md.

Boundaries and Notes

Best for Markdown wikis, knowledge catalogs, and operation logs; not a replacement for database search, access control, or complex knowledge graphs. index.md and log.md must keep stable formats to support grep, awk, or script parsing. log.md grows over time and needs an archival strategy; in multi-user work, append logs in Git instead of overwriting existing entries.