Every past session, subagent, and workflow — searchable in natural language.
Ask in plain language. The agent writes the query, runs it, answers.
---
## What you can ask
```
/obelisk 上次那个 auth 的 bug 我怎么修的
/obelisk 哪些文件这周被反复修改
/obelisk 最近 workflow 跑出来什么结果
/obelisk 我让 subagent 做过哪些代码 review
```
Anything Claude Code has done before — sessions, tool calls, subagents, workflows — is indexed and searchable. Ask in your own words.
## Install
```bash
npx skills add tommy0103/obelisk
```
Or manually: copy `obelisk/` into your project's `.claude/skills/`.
Then in any Claude Code session:
```
/obelisk
```
First run builds the index (~5 seconds for 100 sessions). After that it rebuilds incrementally.
### Requires
- Node.js 22+ (uses built-in node:sqlite with FTS5)
- Claude Code with skills support.
## What gets indexed
| Layer | Source | What's captured |
|-------|--------|----------------|
| **Sessions** | `/.jsonl` | Title, project, timestamps, git branch |
| **Messages** | user + assistant turns | Full text, model, token usage, parent chain |
| **Tool calls** | every tool invocation | Tool name, input, file paths touched |
| **Subagents** | `subagents/agent-.jsonl` | Agent type, description, full conversation |
| **Workflows** | `workflows/wf_.json` | Script, structured result, agent count |
| **Workflow agents** | `subagents/workflows/wf_/` | Per-agent transcripts linked to workflow |
Full-text search via FTS5 covers all message text across every layer.
## How it works
```
You ask a question
↓
Agent writes a JS query against the SQLite index
↓
Runs it via node runtime.mjs --query
