Some intelligence is buried deep in the shadows of the internet – and some of it is built by you, for you, with a few lines of plain English.
This episode covers Issues 105 and 106 of The OSINT Newsletter and explores two very different sides of modern investigations: going deeper into the Dark Web for intelligence, and using AI to automate the repeatable parts of your workflow.
In Episode 18 of The OSINT Podcast, host Jake Creps picks up where Part One left off and takes investigators further into Dark Web intelligence – or DARKINT. He recaps the layered structure of the surface, deep, and dark web, then digs into the beginner toolkit: onion browsers, hidden services, public leak indexes, and onion search engines. Jake walks through a practical methodology for tracing identifiers – emails, usernames, and phone numbers – from breach data on the Dark Web back up to the surface, and explains how to correlate and validate findings without falling for false positives.
He also pulls no punches on the limitations. DARKINT is an adversarial, high-risk environment full of manipulated datasets, unverifiable attribution, and content that can stay with an investigator long after the browser is closed. The episode covers the compliance considerations of handling breached data and the psychological risks of working in this space – and why both deserve serious thought before diving in.
From there, the episode shifts to something brand new: building Claude Skills for OSINT. Jake explains what Claude Skills are – reusable, plain-language instruction sets that turn AI into a reliable part of your investigative workflow – and walks step-by-step through creating one for username search. No code required. He covers picking the right workflow to automate, writing the skill itself, testing it inside Claude Code with Sherlock, and refining it with simple natural-language tweaks.
The episode closes with a look at how to supercharge an OSINT skill: instructing Claude to pivot on its own findings, find new tools when collection hits a wall, and fall back to manual methods when scripts fail. It is a glimpse of what investigative automation actually looks like when AI stops being a novelty and starts doing real work alongside the analyst.
Highlights include:
🧅 DARKINT Toolkit – onion browsers, hidden services, leak indexes, and onion search engines explained for beginners curious about web spelunking.
🔗 Surface-Bound Pivots – a step-by-step methodology for tracing emails, usernames, and phone numbers from breach data back to the surface web.
⚠️ The Monsters Under The Bed – the real limitations of DARKINT, from manipulated datasets and unverifiable attribution to the psychological toll of the work.
🤖 Building Claude Skills – how to turn a repeatable OSINT workflow into a reusable Claude Skill, with a full walkthrough of automating a username search using Sherlock.
🚀 Supercharging Automation – instructing Claude to pivot on its findings, hunt for new tools, and fall back to manual methods when scripts come up short.
Whether the data is hiding in the dark or waiting to be unlocked by the right set of instructions, Episode 18 shows how modern investigators are reaching both.
References
