Over the past few weeks I've been building OSINT Navigator — a RAG-powered search engine that consolidates 7,500+ OSINT tools into a single interface, with an MCP for agents to use directly in the CLI. Building this with craig and alexios at Indicator to launch it alongside experts in the field, to a large community of open-source researchers.

The database of OSINT tools we've pulled from online toolkits is open-source, with sources accredited for each row.

Alongside Navigator, I've also built an OSINT skill file — a structured markdown reference that gives CLI agents and local models investigative expertise to draw on.

The OSINT skill

The skill gives you three capabilities:

/osint — tool catalog

/investigate — methodology

/follow-the-money — financial investigation

The skills budget and revenue methodology was built by extracting structured data with LlamaParse from the Open Society Institutes Follow the Money: A Guide to Monitoring Budgets and Oil and Gas Revenues by Jim Shultz (Revenue Watch, 2005).

The three work together: /osint points to /investigate when you need technique, not just a tool, and /investigate points back when you need alternatives. Online, both tap into OSINT Navigator — a database of 7,500+ tools with country-specific sources and weekly updates — turning a three-tier system (tool catalog → methodology → deep discovery) into a single workflow. Both skills also work fully offline if you need your queries to stay on your machine.

The OSINT community runs on shared knowledge. We wanted to make these tools available to anyone, with options to use them at no cost.

Install via Claude Code

If you use Claude Code, install directly from the plugin marketplace:

/plugin marketplace add buriedsignals/skills
/plugin install osint@buriedsignals

Once installed, the skill activates automatically when you ask investigation questions.

Using it with LM Studio

OSINT requires being online. But if you'd rather keep your investigation questions off a frontier model API, you can run the skill files on a local model with LM Studio.

Load a model, larger models handle multi-step workflows better — and use LM Studios Chat with Files feature. I suggest downloading these reference files, attaching them to your chat, and asking your questions:

LM Studio injects the full file contents into the models context. No chunking, no embedding loss. Your questions stay on your machine.

LM Studio with Gemma 3 4B running the OSINT skill via Chat with Files gemma 3 4B in LM Studio with the OSINT reference files attached via Chat with Files.

Links