[Trend Scout] kitepon-rgb/Caveat

[github-actions]

🔭 Trend Scout: kitepon-rgb/Caveat

Auto-generated by trend-scout.py — review and edit as needed.

📌 What problem it solves

Long-term memory layer for Claude Code and Codex — markdown-in-git knowledge base, SQLite FTS5 retrieval, MCP, and native agent hooks.

📅 Timeline

Field	Value
Created	2026-04-18
Last pushed	2026-07-17
Stars	2
Forks	0
Open issues	7
License	MIT
Language	TypeScript
Topics	agent-tools, ai-tools, anthropic, claude-code, cli, codex, coding-agents, developer-tools, fts5, knowledge-base, markdown, mcp, obsidian, openai, sqlite, typescript

✅ Strengths

• Well-tagged: agent-tools, ai-tools, anthropic, claude-code, cli, codex
• Primary language: TypeScript
• Actively maintained (pushed within 30 days)

⚠️ Weaknesses / Risks

• Low star count (2) — limited community validation

💡 What this repo can learn

• Claude Code session patterns: this repo's Claude Code integration approach could improve claude-adapter.py's JSONL parsing — e.g., handling new session event types or extracting richer metadata from Claude Code tool-use blocks
• CLI verb patterns: a clear add/search/update/delete verb model (like memory-tool add / search / dream) could streamline the UX of query-session.py and learn.py, making them easier to invoke from hooks or scripts
• Git hook / workflow patterns: hook design from this repo could strengthen the hooks/ enforcement chain (e.g., auto-briefing, commit guards, learn reminders)
• MCP tool-server surface: the MCP server interface here could expose query-session.py and briefing.py as directly callable MCP tools — e.g., a Copilot agent could invoke the briefing lookup in-process rather than shelling out to a subprocess, reducing latency in hook-driven workflows

README excerpt

<p align="center">
  <img src=".github/og.png" alt="Caveat — long-term memory layer for coding agents" width="100%">
</p>

# Caveat

[![npm](https://img.shields.io/npm/v/caveat-cli?color=cb3837&label=caveat-cli)](https://www.npmjs.com/package/caveat-cli)
[![CI](https://github.com/kitepon-rgb/Caveat/actions/workflows/ci.yml/badge.svg)](https://github.com/kitepon-rgb/Caveat/actions/workflows/ci.yml)
[![license](https://img.shields.io/npm/l/caveat-cli?color=blue)](LICENSE)
[![node](https://img.shields.io/node/v/caveat-cli?color=339933&logo=node.js&logoColor=white)](https://nodejs.org/)
[![GitHub release](https://img.shields.io/github/v/release/kitepon-rgb/Caveat?color=24292e&logo=github)](https://github.com/kitepon-rgb/Caveat/releases)

> **Stop rediscovering the same trap.** Caveat is a long-term memory layer for Claude Code and Codex: every time you bleed for an external-spec quirk or a repo-specific oddity, write it down once — and the next time anyone (you or your AI) is about to step on the same rake, the relevant note surfaces automatically.

🇯🇵 **日本語版**: [README.ja.md](README.ja.md)

## What it does in 30 seconds

```sh
npm install -g caveat-cli
caveat init                          # registers Claude Code MCP + hooks
caveat codex-hook install            # optional: register native Codex hooks

On macOS with Homebrew Node, generated hook commands use the stable
/opt/homebrew/bin/node symlink when it resolves to the current Node binary,
instead of a versioned `/opt/ho

(truncated)

</details>

---
*Scouted on 2026-07-17 · [View on GitHub](https://github.com/kitepon-rgb/Caveat)*

<!-- trend-scout:repo:c8cb7d8d79cc44a2 -->