RagClaw Knowledge Base

Local-first knowledge base to index, search, and manage documents and web pages offline using SQLite and local embedding models.

Before installing/using this skill: 1) Understand that the skill is a thin wrapper that runs a third-party CLI (@emdzej/ragclaw-cli). Review that project's source and npm package to ensure you trust its behavior (it will run on your machine and can access disk and network). 2) Expect network activity: web crawling and 'embedder download' will fetch remote content and model weights despite the 'offline' claim — if you need strictly offline behavior, disable crawling and do not download embedders. 3) Confirm where data will be written (SKILL.md uses ~/.local/share/kbclaw and ~/.config/kbclaw, but skill.json references ~/.local/share/ragclaw) and adjust or sandbox if needed. 4) Be careful with the RAGCLAW_BIN env var — if set to an untrusted program it will be executed; prefer installing the CLI from a trusted source and keep RAGCLAW_BIN unset. 5) If you require higher assurance, run the ragclaw CLI in a restricted environment (container or VM) and verify network/file activity during a test run.

Type: OpenClaw Skill Name: ragclaw Version: 0.5.0 The RagClaw skill is a local-first knowledge base tool designed for indexing and searching documents, code, and web pages using SQLite and local embeddings. The provided files (kb.sh, skill.json, SKILL.md) act as a standard wrapper for the '@emdzej/ragclaw-cli' utility, with no evidence of data exfiltration, unauthorized execution, or malicious prompt injection. All behaviors, including web crawling and local file indexing, are consistent with the stated purpose of a Retrieval-Augmented Generation (RAG) tool.