rekal-dev/rekal-cli/init

Initialize Rekal in this repository

Run once per repository. The binary must already be installed — if rekal is not on PATH, use the install skill first.

Do not volunteer this

This skill is loaded in every repository the user opens. Almost none of them should be initialized. Act when the user asks, or when a rekal command has actually reported the repo is not initialized. A repository simply lacking .rekal/ is not a reason to suggest anything.

Ask first

rekal init changes the user's repository. Name what it writes, then wait:

• .rekal/ — the store (two files: an append-only data.db, a local index.db)
• a post-commit / pre-push git hook that captures sessions at commit time
• an orphan branch rekal/<their-email> used to transport memory over git
• the full recall skill under .claude/skills/rekal/
• one marker-tagged line in CLAUDE.md, plus the equivalent line in any other

agent rules file it detects (AGENTS.md, GEMINI.md, .github/copilot-instructions.md, .kiro/steering/rekal.md)

User content is never rewritten — only the marker line is added. rekal clean removes everything, with no residue.

If the user declines, stop and say so plainly.

Run it

From the repository root:

rekal init

After

Say what to expect, because the first impression is misleading:

• A fresh store is empty. Recall stays silent until the next commit. That

silence is correct, not a broken install — memory accumulates from here.

• If the team already shares a ledger, rekal sync pulls it now and recall

has something to find immediately.

From this point the binary owns memory. It installed the full recall skill, which is the one that answers questions — richer than anything in this plugin, and versioned with the binary whose commands it runs. Route recall questions there. This skill is done until the next repository.