Claude Guide

Claude uses the same provider-neutral hook runtime as Codex. Same cards, same retrieval engine.

Supported mapping

Claude event	Normalized event
UserPromptSubmit	prompt.submit

Hook installation

ome init --provider claude --dry-run   # Preview
ome init --provider claude             # Install

The installer writes to ~/.claude/settings.json.

Have your agent do it:

Help me install the Oh My Experience Claude hook.

1. Run `ome init --provider claude --dry-run` first to preview what will be written.
2. If the preview is safe, run `ome init --provider claude`.
3. Run `ome hook status --provider claude` to confirm the hook is enabled.

Using both Codex and Claude

If you use both:

ome init --provider all

One library, one retrieval engine, shared automatically. No duplicate rules.

Running a reflect scan with Claude

Copy and paste this prompt:

Run an OME reflect scan over my recent coding sessions.

1. Run ome reflect start --focus "recent execution mistakes I corrected"
2. Check recent conversations for places where I corrected you
3. Generate ≤5 candidates in the current OME candidate JSON shape: audit plus candidates with summary, criteria.use_when, criteria.ignore_when, recall, optional engine_hints, scope, and rule.
4. Write candidates to candidates.json, then ome reflect candidates RUN_ID --from-file candidates.json
5. ome reflect show RUN_ID to display candidates. Wait for my approval on each one.

Only extract reusable execution judgment. Don't turn one-off context into experience cards.

Rule

Do not fork retrieval logic for Claude. The Claude adapter only handles hook input, install/status paths, and output formatting.