Claude Lore Skill — Documentation skill for Claude Code

claude-lore-skill

A Claude Code skill + rulebook addition for keeping a `LORE.md` chronicle in your project — the *why we are here* alongside `CHANGELOG.md` (*what shipped*), `CLAUDE.md` (*how it works*), and `TODOS.md` (*what's next*).

The chronicle answers questions other docs can't: *why did we choose this? Who was in the room when this got decided? What did we almost do instead?* Future readers — collaborators, agents, future-you — pick up the project with the texture intact.

Why this exists

CHANGELOG tells you what shipped. CLAUDE.md tells you how the code works. TODOS tells you what's next. None of them tell you the *why* — the conversations, decisions, tradeoffs, and shifts that produced the current state. After a few months on a project, the why is gone unless someone wrote it down.

`LORE.md` is where the why lives. This repo gives you two things to keep it alive:

1. A rulebook addition (`lore/references/upkeep-rule.md`) — paste it into your CLAUDE.md or AGENTS.md and your agent will run a LORE check on every change-set landing on main.
2. A chronicler skill (/lore) — invoked manually when you want to read or enrich the chronicle, or invoked proactively by the rulebook at substantive moments (version bumps, multi-module changes, long-running PRs landing) to ask one grounded journalist-question about what was behind the change.

The two work together: the rule keeps the chronicle fresh; the skill makes it richer.

Install

1. The skill

git clone https://github.com/deniz-miro/claude-lore-skill.git
cp -r claude-lore-skill/lore ~/.claude/skills/lore

The skill is now available as `/lore` in any Claude Code session.

2. The rule

Open [`lore/references/upkeep-rule.md`](lore/references/upkeep-rule.md) and paste its contents into your project's `CLAUDE.md` or `AGENTS.md`. From the next change-set forward, your agent will run the LORE check at the right moments. (When you invoke `/l