Skill flagged — suspicious patterns detected
ClawHub Security flagged this skill as suspicious. Review the scan results before using.
Local Rag Search 0.1.0
v1.0.0Efficiently perform web searches using the mcp-local-rag server with semantic similarity ranking. Use this skill when you need to search the web for current...
⭐ 0· 62·0 current·0 all-time
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
OpenClaw
Suspicious
high confidencePurpose & Capability
The skill claims to enable RAG-based web search via an mcp-local-rag MCP server and the SKILL.md consistently documents the rag_search_* and deep_research tools. It requires no env vars or binaries itself — but it depends on the external mcp-local-rag server being installed. That dependency is coherent with the skill's stated purpose, but the skill does not bundle or install the server itself; the README instructs users to install code/images from GitHub/GHCR which is an out-of-band step the user must trust.
Instruction Scope
SKILL.md stays within search-related actions: choosing backends, forming queries, tuning parameters, citing URLs, and switching backends on error. It does not instruct reading system files or environment variables. Minor inconsistencies: examples reference 'stackoverflow' as a backend though it's not listed among supported backends, and the skill alternately claims 'no external APIs' and 'all processing runs locally' while still describing queries against Google/Bing/etc. (these imply outbound network requests via the MCP server).
Install Mechanism
The skill itself is instruction-only (no install spec), but README provides explicit install snippets that pull code via 'git+https://github.com/...' with uvx or by running a Docker image from ghcr.io/nkapila6/mcp-local-rag:v1.0.2. Pulling and running code or containers from external repositories is a non-trivial trust decision — while GitHub and GHCR are common release hosts, you should review the upstream repository and container image before running them locally.
Credentials
The skill declares no required environment variables, no credentials, and no config paths. The runtime instructions do not request secrets or unrelated credentials. This is proportionate for a client-side search helper that relies on an MCP server.
Persistence & Privilege
Skill flags are default (always:false, user-invocable, agent invocation allowed). The skill does not request persistent system-wide privileges or attempt to modify other skills. No concerning persistence behavior is present in the package contents.
What to consider before installing
This skill is mostly instruction-only and appears to teach an agent how to use a local MCP 'mcp-local-rag' server to perform semantic web searches. Before installing or enabling it: 1) verify and review the upstream mcp-local-rag repository and the GHCR Docker image referenced in README (pulling/running containers or installing code from GitHub executes third-party code). 2) Note metadata inconsistencies (version/owner IDs between registry metadata, _meta.json, and package.json); ask the publisher to clarify if provenance matters. 3) Be aware searches performed by the MCP server will fetch external web URLs — if you need strict local-only behavior, confirm the server runs in an isolated/local environment. 4) If you want to proceed, run the MCP server and container in a sandboxed/test environment first and inspect the repo/image contents (or build from source) to reduce risk.Like a lobster shell, security has layers — review code before you run it.
latestvk978y88x3cx71m0v1m8vq41z1n83q9c0
License
MIT-0
Free to use, modify, and redistribute. No attribution required.