Memory Wiki — AI skill for Claude Code

memory-wiki

**Part of the Evidence-first Agents suite** — tooling that makes AI agents accountable instead of just capable: [answer-contract](https://github.com/chenhz01/answer-contract) (output discipline) · [skill-spec](https://github.com/chenhz01/skill-spec) (spec discipline) · [memory-wiki](https://github.com/chenhz01/memory-wiki) (memory discipline). Same author, same zero-dependency philosophy.

**The file-layer wiki for your agent's markdown memory.** Your agent remembers in scattered `.md` files. Give those files an index, a search, and a health check — without a database, a server, or a single dependency.

Agent memory tools keep racing toward vector stores and hosted services. But most agents (and most people) keep memory where it already works: **plain markdown files**. memory-wiki is the missing file layer — it treats your memory folder like code: indexed, searchable, and linted.

Inspired by the persistent-wiki pattern (see [nashsu/llm_wiki](https://github.com/nashsu/llm_wiki)); this is an independent, dependency-free implementation. Pairs with [skill-spec](https://github.com/chenhz01/skill-spec) — specs for your agent's skills, a wiki for its memory.

What it does

Command	Effect
ingest	Absorb markdown files as entries (auto-dated if undated, slugified, never overwrites without --force)
index	Rebuild index.md — one table: title, date, first line, linked
query	Keyword search across entries, ranked by hit count, with context lines
lint	Health check: orphans not in the index, missing dates, duplicate titles, overlong files, broken relative links
log	Append-only operation log (log.md) — what entered the wiki and when

Install (60 seconds)

Single file. Zero dependencies. No server, no database, no network calls — your memory never leaves your disk.

git clone https://github.com/chenhz01/memory-wiki.git
python memory-wiki/tools/memory_wiki.py ingest my-notes.md --wik