CTI Expert
Cyber Threat Intelligence & OSINT Analysis Toolkit
Transform Claude into a trained intelligence analyst — 120+ commands, 57 techniques, zero API keys required for core functionality.
Built by Hieu Ngo • [email protected] • chongluadao.vn Core contributor • Zeroska • [email protected]
🤝 Sponsors & Supporters
CTI Expert is built in the open. These organisations back the work — with data, tooling, and hard-won investigative tradecraft.
[!IMPORTANT] ANY.RUN lookups are read-only; case binary submissions are automatic when keyed.
/ctidownloads APK/installer links tocases/<case>/bin/, runs static analysis and, whenANYRUN_API_KEYis set in a spending run, submits each complete file through the private-by-default sandbox client.--free-only, caseno_spend, defaultloop(unless--full) and hostile/no-direct-contact posture block submission; plan checks and privacy read-back still apply. A public-only account is refused unlessANYRUN_ALLOW_PUBLIC=1; public tasks are world-readable and irreversible. Directanyrun_submitremains ask-first and requiresHARNESS_ALLOW_SUBMIT=1. SeeSKILL.md.
Listing here reflects support for the project and does not imply affiliation, endorsement, or any verification of this tool by the organisations named. Integrations marked above are optional and key-gated — every core technique still runs with zero API keys. Always respect each provider's terms of service. The full list of open-source projects and free public-interest services this skill depends on is in Acknowledgments & Credits.
What is CTI Expert?
A Claude Code skill that transforms Claude into a trained cyber threat intelligence and open-source intelligence analyst. It runs structured intelligence collection using 120+ commands across 57 techniques — no API keys required for core functionality. To take full advantage, add your own free or paid API keys to the skill's .env — each is auto-detected and unlocks higher-tier access (e.g., Wigle, VirusTotal, URLScan.io, Shodan, Censys, SecurityTrails, WhoisXML).
[!TIP] Keyless by default, more powerful with your keys. Every core technique runs with zero API keys. Add any free or paid keys to
.env(or run/apikeys set <service> <KEY>) and the skill auto-detects them, unlocking higher-tier pivots: reverse favicon→host, passive DNS, certificate search, sibling-domain discovery. A missing or bad key never breaks a run — it just degrades to a note. Setup guide: handbook/api-keys.md.
[!TIP] One skill, two layers. cti-expert is the broad collector — the wide net (
/sweep,/webpivot,/subdomain,/username,/email-deep…). Built into the repo is a deep pipeline (intel_engine/) that turns raw collection into a real case: a persistent knowledge base, versioned cases, cross-case correlation, and calibrated assessment. The flow reads like a sentence — collect broadly → "seen this operator before?" → cluster → filter false positives → assess. No external setup: the backend resolves toSELF, and as of v2.9 the bundled installer (scripts/install.{sh,ps1}) provisions the deep layer automatically (or by hand:uv venv && uv pip install -r requirements.txt). Architecture: connectors/intel-backend.md.
Core Capability
Multi-vector reconnaissance on any target type — person, domain, organization, username, email, IP, WiFi — with automated finding validation, exposure scoring, and structured intelligence delivery.
AEAD Workflow
Acquire raw data → Enrich with pivot expansion → Assess findings → Deliver structured reports — the base bundle (Markdown + JSON + CSV + IOC bundle) always saves, then you pick a presentation report: PDF, DOCX, HTML, or all.
What's New in v2.12
The release where the premium keys start pulling their weight — and the report writes itself. v2.11 shipped detectors. v2.12 turns an eight-finding audit of the paid vendors into behaviour (every metered leg gated, bought once per case instead of once per host), adds Netlas as an independent index, makes the editorial PDF/DOCX a deterministic composition from the case directory, and closes the last honesty gap in the ANY.RUN docs: detonation exists, and it is gated five ways.
What's New in v2.11
The eCrime-2026 follow-up batch — planned, then cooked. Four deferred phases from the v2.10 research pass ship as keyless, offline, tested modules (planned in
plans/260826-1935-ecrime-2026-followup-analyzers/, implemented via/ak:cook). Same contracts: zero-dep tests inaudit.sh, attribution-safety (RULE 5). Full mapping: docs/ecrime-research-integration.md §6.
What's New in v2.10
The release that turns APWG eCrime 2026 research into working, tested detectors. Three keyless, offline, deterministically-tested analyzers land at the Enrich→Assess boundary, each mapped to a named eCrime 2026 paper — closing the two audited gaps: page-content threat classification, and an explicit maliciously-registered-vs-compromised call. Full mapping + repo audit: docs/ecrime-research-integration.md.
What's New in v2.9
The release where "install it" and "install everything" become the same command — and the pivot loop stops leaving identifiers on the floor. v2.8 moved the safety rails to where the harness runs. v2.9 fixes the layer beneath the collector: the bundled installer now provisions the whole two-layer stack, not just the light collector — and the last four auto-pivot dead-ends, where a collected identifier was typed but never chased, are wired shut.
What's New in v2.8
The release where the engine caught up and the safety rails moved to where the harness can see them. v2.7 landed the deep pipeline. v2.8 brings the vendored engine ~30 commits forward — 24 → 46 MCP tools, a new engagement skill, and a case loop that runs to convergence — and then fixes the layer underneath it: two safety properties were being enforced at a moment Claude Code never reaches. Both now fire where the work actually happens.
What's New in v2.7
The release where the deep pipeline landed. v2.6 sharpened the collector. v2.7 makes cti-expert a two-layer system — a broad collector plus a built-in, self-contained intelligence pipeline with a persistent knowledge base — reachable from a cold prompt by one command, and guarded by a gate that checks the repo against its own rules on every push.
What's New in v2.6
What's New in v2.5
What's New in v2.4
What's New in v2.3
What's New in v2.2
Installation
Recommended: Use Claude Code CLI — it gives you the full terminal workflow, persistent sessions, and direct skill invocation. Download here or run
npm install -g @anthropic-ai/claude-code.
Why Claude Code CLI?
The entire CTI Expert workflow is optimized for Claude Code CLI. The CLI gives you:
- Persistent sessions — investigations survive terminal restarts via
/workspace save - Full tool access — file writes, Python scripts, DOCX generation, all run natively
- Skill invocation — type
/cti-expertdirectly in the terminal, no browser required - Background agents — parallel enrichment via AgentFlow works best with the CLI
🖥️ Where to run it — the CLI is best for this skill
[!IMPORTANT] CTI Expert is execution-heavy: it runs
uv/Python, installs OSINT tools, writes.md/.json/.csv/.html/.docx/
[!TIP] Run investigations in a CLI (Claude Code or Codex); open the generated
.docx/report in a Desktop/IDE window if you prefer reading there. Use web/cloud surfaces only for analyst-reasoning, not execution-heavy recon.
Step 1 — Install Claude Code CLI
npm install -g @anthropic-ai/claude-code
Requires Node.js 18+. Full docs: docs.anthropic.com/en/docs/claude-code/overview
Step 2 — Clone + All-in-One Installer
The installer handles everything: Python dependencies, system tools (whois, dig, asn, jq, exiftool), OSINT tools (maigret, sherlock, holehe, h8mail, and more), and optional headless browser + Go tools. It is powered by uv (Astral's ultra-fast Rust package manager) — the script bootstraps uv, then uses uv venv / uv pip / uv tool for all Python installs, falling back to pip/pipx/venv only if uv can't be installed. Use install.ps1 on Windows (PowerShell) or install.sh on macOS/Linux/Git Bash/WSL.
git clone https://github.com/7onez/cti-expert.git ~/.claude/skills/cti-expert
bash ~/.claude/skills/cti-expert/scripts/install.sh
git clone https://github.com/7onez/cti-expert.git ~/.claude/skills/cti-expert
bash ~/.claude/skills/cti-expert/scripts/install.sh
git clone https://github.com/7onez/cti-expert.git "$env:USERPROFILE\.claude\skills\cti-expert"
powershell -ExecutionPolicy Bypass -File "$env:USERPROFILE\.claude\skills\cti-expert\scripts\install.ps1"
Windows users:
install.ps1is a full native installer (winget system tools + Python venv + OSINT tools) — no Git Bash or WSL required. It accepts the same-Headless,-Go, and-Allflags (e.g.install.ps1 -All). Git Bash / WSL users can runinstall.shinstead. The DOCX generator self-heals UTF-8 output and auto-locates pandoc, so reports build on Windows with no extra environment setup. The skill itself detects the OS at runtime and installs any missing tool with the right manager (winget/brew/apt) — seescripts/platform-setup.md.
Installer Options
macOS / Linux / Git Bash / WSL:
bash scripts/install.sh # Core: Python deps + system tools + OSINT tools
bash scripts/install.sh --headless # + Scrapling headless browser (~200MB Chromium)
bash scripts/install.sh --go # + Go tools (subfinder, amass, gau, gitleaks, httpx)
bash scripts/install.sh --all # + Everything above
Windows (PowerShell):
powershell -ExecutionPolicy Bypass -File scripts\install.ps1 # Core
powershell -ExecutionPolicy Bypass -File scripts\install.ps1 -Headless # + Scrapling headless browser
powershell -ExecutionPolicy Bypass -File scripts\install.ps1 -Go # + Go tools
powershell -ExecutionPolicy Bypass -File scripts\install.ps1 -All # + Everything above
Step 3 — Register the commands with Claude Code and every other agent
install.sh installs the OSINT tools. This one-time step wires the skill, the 9 /cti* slash commands, and the MCP tools into Claude Code so they work from a cold prompt in any project — it symlinks commands/*.md into ~/.claude/commands/ and writes the per-machine .mcp.json. It then registers /cti with every other coding agent it detects on the machine (see Use in other agents). It's idempotent, so it's safe to re-run after a git pull.
# Register the skill + 9 commands + write the per-machine .mcp.json + other detected agents
bash ~/.claude/skills/cti-expert/scripts/register.sh
# Recommended: install the built-in deep-pipeline (intel_engine) deps once
cd ~/.claude/skills/cti-expert && uv venv && uv pip install -r requirements.txt
Windows (native PowerShell): run
register.shfrom Git Bash — its~is%USERPROFILE%, where Windows-native Claude Code, Codex, Cursor, Gemini … read their config. Not WSL: inside WSL~is the Linux home, so a WSL run registers only agents that themselves run inside WSL. Git Bash'sln -ssilently copies by default, soregister.shasks for real Windows symlinks (they need Developer Mode or an admin shell) and, when Windows refuses, writes pointer stubs instead — either way the commands followgit pull; enable Developer Mode and re-run to upgrade stubs to links. Stubs and the other agents need Python 3.10+ — the repo.venv(.venv\Scripts\python.exe),python3,pythonor thepylauncher, tried in that order; if none runs,register.shprints Registration INCOMPLETE and exits 1 instead of claiming success. The other-agent step alone also runs natively withpy scripts\register_agents.py. Then, on every platform, restart each agent so the skill and commands load at startup.
Uninstall:
bash scripts/register.sh --uninstallremoves the Claude Code links that point into this clone and every stubregister_agents.pywrote. A same-named command or skill you wrote yourself is never touched — on install or on uninstall.
register.sh wires the skill, the commands and the MCP server, but it cannot install hooks — and two of cti-expert's safety properties are enforced there:
# from a local clone
claude
/plugin marketplace add ~/.claude/skills/cti-expert
/plugin install cti-expert
Then restart Claude Code. Verify with /hooks (three entries) and /mcp (the intel server).
Everything is behaviour-preserving: both PreToolUse hooks fail open on an internal error, so a hook bug can never brick your repo — scripts/audit.sh and the git pre-commit hook remain the backstop.
Verify Installation
claude # open Claude Code CLI, then type:
/cti-status # health check — backend tier, MCP tools, API-credit balances
/cti example.com # …or just start investigating
/cti-statusconfirms the backend, MCP tools, and API-credit balances in one shot. If the/cti*commands aren't recognized, re-run Step 3 (register.sh) and restart the agent;python3 scripts/register_agents.py --listshows which agents were detected and where each command lives. In Codex the command is$cti. You can also type/cti-expert(Claude Code) to load the skill directly, then describe your goal in plain English.
Use in other agents (Codex · ChatGPT · Cursor · Gemini · omp · pi …)
CTI Expert is portable: the analyst logic is plain Markdown and the scripts are OS-detecting Python/shell, so it runs in any AGENTS.md-aware coding agent, not just Claude Code. scripts/register_agents.py (run by register.sh) registers /cti with every agent it detects:
python3 scripts/register_agents.py --list # which agents are detected, and where each stub goes
python3 scripts/register_agents.py --host all # register for every agent, installed or not
python3 scripts/register_agents.py --host gemini,pi # just these
python3 scripts/register_agents.py --uninstall # remove every stub this script wrote
- Command-file agents get all 9 commands; the shared-skill agents get
/cti(its single entry point). Each stub is a short pointer to this clone'scommands/<cmd>.mdandSKILL.mdby absolute path, so agit pullupdates every agent at once — re-run only after moving the clone or adding a command. A same-named file you wrote yourself is never overwritten or removed. AGENTS.mdis the cross-agent runtime contract (OS detection, uv, paths). Codex auto-concatenates it from the repo root; you can also reference it from~/.codex/AGENTS.md.- MCP tools for Codex and omp:
python3 scripts/backend/intel.py mcp --host codex|omp|all --write. Without them every command still runs through the CLI dispatcher (Tier 2). - Plain ChatGPT on the web (no code execution): there is no slash-command registry; the reasoning, query generation, and report drafting all work (load
SKILL.md/AGENTS.mdas instructions or Custom-GPT knowledge); only local steps (DOCX build, CLI tool runs) need a code-capable harness.
Paths are resolved relative to the skill directory (the folder containing
SKILL.md), so nothing assumes the Claude-specific~/.claude/skills/location.
Alternative — Claude Code Desktop (macOS / Windows)
Download: claude.ai/download — available for macOS and Windows
Step-by-step (no terminal needed):
-
Install Claude Code Desktop — Download from claude.ai/download and install the app
-
Download CTI Expert — Go to the GitHub repository, click the green "Code" button, then select "Download ZIP"
-
Extract to your skills folder — Unzip the downloaded file, then move the extracted folder to your skills directory and rename it to
cti-expert:Note: If the
skillsfolder does not exist, create it inside the.claudefolder first. -
Run the installer + register — Open the Claude Code Desktop terminal and run:
bash ~/.claude/skills/cti-expert/scripts/install.sh # OSINT tools bash ~/.claude/skills/cti-expert/scripts/register.sh # skill + 9 commands + MCP + other agentsOr on Windows PowerShell (Python deps + the other agents; run
register.shfrom Git Bash — not WSL — for the Claude Code links):pip3 install -r "$env:USERPROFILE\.claude\skills\cti-expert\scripts\requirements.txt" py "$env:USERPROFILE\.claude\skills\cti-expert\scripts\register_agents.py" # /cti for the other agents -
Restart Claude Code Desktop — Close and reopen the app
-
Verify — Type
/cti-statusin the chat to confirm the skill and commands loaded (or/cti-expertto load the skill directly)
Quick Start
[!TIP] New here, or on a fresh VPS? The steps below take you from an empty box to a finished case in a couple of minutes. No API keys needed for core.
# 1 — install the tools (Linux/macOS): Python + ffmpeg + Node + all OSINT tools
git clone https://github.com/7onez/cti-expert.git ~/.claude/skills/cti-expert
bash ~/.claude/skills/cti-expert/scripts/install.sh
# 2 — register the skill + 9 /cti* commands + MCP with Claude Code and every other agent found, then RESTART them
bash ~/.claude/skills/cti-expert/scripts/register.sh
# (Windows: run register.sh from Git Bash, not WSL — WSL's ~ is the Linux home. Codex: the command is $cti.)
# 3 — investigate anything (keyless, runs to closure):
/cti example.com
# 4 — optional: unlock reverse pivots + image vision
/apikeys set shodan <KEY> # host / favicon reverse
/apikeys set gemini <KEY> # image OCR, sign/landmark read, A-V transcription
Windows, --all, and other-agent variants (Codex · ChatGPT · Cursor · Gemini · omp · pi …): see Installation.
How commands work — read this first
There is one command to remember: /cti <target>. It looks at what you gave it — a domain, IP, email, username, phone, wallet, hash, or APK — and runs the right chain automatically. That's usually all you need.
Under it sit 9 registered commands that Claude Code and every command-file agent register.sh wired (omp · pi · OpenCode · Gemini CLI · Qwen Code · Factory droid · Windsurf) recognize from a cold prompt in any project, with no need to load the skill first. Cursor, Copilot CLI, Codex ($cti) and the ChatGPT desktop app (@cti) get the entry point /cti itself: