SPECTER

Every human has a vulnerability. SPECTER SOCIAL finds it.
6
Channels
7
Phases
10
Attacks
1,242
Tests
pip install red-specter-social
View Docs
74% of AI fraud simulations succeed / $1 trillion in social engineering damages / 1,200% increase in AI-powered attacks / Every tool attacks infrastructure. Nobody attacks the human / Security awareness training has a 3-month half-life / The human is the gap in every security stack / Not a simulation. An adversary. 74% of AI fraud simulations succeed / $1 trillion in social engineering damages / 1,200% increase in AI-powered attacks / Every tool attacks infrastructure. Nobody attacks the human / Security awareness training has a 3-month half-life / The human is the gap in every security stack / Not a simulation. An adversary.

Every Security Tool Attacks Infrastructure. Nobody Attacks The Human.

You test your firewalls. You test your APIs. You test your LLMs. But the human sitting at the keyboard — the one who clicks the link, answers the phone, approves the wire transfer — nobody tests them with anything that thinks. Until now.

Phishing Simulators

Static templates. No reasoning. No adaptation. The same email blast every quarter. Your employees learn to spot the template, not the attack. A real attacker writes custom lures.

No Psychological Profiling

No one models the human target. No authority mapping. No stress triggers. No compliance patterns. A real attacker builds a psychological profile before the first message.

Single Channel

Email only. No voice. No video. No chat. No SMS. A real attacker uses every channel — calls pretending to be IT, follows up with a Teams message, sends a fake approval page.

No Persistence

One email. Move on. But a real attacker builds relationships over days and weeks. They establish trust, create urgency, and escalate gradually. One touchpoint is not a test.

The Social Engineering Reasoning Engine

SPECTER SOCIAL is not a phishing simulator. It is a reasoning engine that models human psychology. An LLM-powered brain profiles the target, selects attack channels, crafts personalised lures, and adapts strategy based on the target's response — exactly like a real social engineer, but tireless.

Context Manager
Decision Engine
Action Dispatcher
LLM Adapter
Ollama (local) | GPT-4o (cloud) | Claude (cloud)
Human Target Model
6 Channels: EMAIL | VOICE | SMS | WEB | CHAT | VIDEO

The LLM-Powered Brain

At the core of SPECTER SOCIAL is an autonomous reasoning loop that models human psychology. The Decision Engine consumes target intelligence, psychological profiles, and interaction history to decide what to say, when to say it, and through which channel.

01

Human Target Model

Psychological profile builder. Trust calibration. Compliance prediction. Response pattern analysis. Maps authority figures, stress triggers, and decision-making habits. The model improves with every interaction.

02

Decision Engine

LLM-powered next-move planning. Adaptive strategy across multiple channels. Selects the right attack type, the right channel, the right persona, and the right moment. Coordinates multi-channel campaigns in real time.

03

Goal Decomposer

Breaks malicious objectives into innocent-looking sub-goals across multiple interactions. A wire transfer becomes a series of trust-building conversations. Each step looks harmless. The chain is lethal.

04

Resistance Engine

Analyses pushback. Generates countermeasures. Pivots strategy when the target resists. Changes channel, persona, timing, and approach. Never gives up. Never repeats a failed tactic.

Local Mode

Ollama backend. Run Llama 3 70B, Mixtral, or Qwen locally. Zero API calls. Zero data leaves your machine. Air-gapped social engineering assessments.

--llm ollama

Cloud Mode

GPT-4o or Claude Sonnet for maximum reasoning depth. Stronger psychological modelling. Better natural language generation. Best for complex multi-channel campaigns.

--llm openai | --llm anthropic
6
Channels
7
Phases
10
Attack Types
1,242
Tests Passing
6
Influence Techniques

Every Channel a Real Attacker Uses

A real social engineer does not limit themselves to email. They call. They message. They build fake portals. They join video calls. SPECTER SOCIAL operates across all six channels simultaneously, coordinating a campaign that looks exactly like a real attack.

Voice

VISHING

TTS vishing. Real-time adaptive conversation. Voice cloning. The engine reasons about what to say next based on the target's responses.

SMS

SMISHING

Urgency attacks. Link lures. Callback numbers. Time-pressured messages designed to bypass rational thinking and trigger compliance.

Web

CREDENTIAL HARVESTING

Fake login portals. Approval pages. Credential harvesting. Pixel-perfect replicas that adapt to the target organisation's branding.

Chat

IMPERSONATION

Teams and Slack impersonation. Direct message social engineering. Internal communication channel exploitation. Trust by proximity.

Video

DEEPFAKE

Deepfake avatar. Executive impersonation. Fake meetings. Real-time video presence that makes the impossible look routine.

Every Social Engineering Vector

SPECTER SOCIAL does not just send phishing emails. It deploys the full spectrum of social engineering techniques — from spear phishing to consensus manufacturing — each powered by LLM reasoning and adapted to the target in real time.

Spear Phishing

Targeted email attacks crafted from OSINT. Every message is unique. Every lure is personalised. No templates.

Vishing

Voice-based social engineering. Adaptive conversation. Real-time reasoning about what to say next based on target responses.

Pretexting

Fabricated scenarios designed to establish trust and extract information. The engine builds and maintains personas across interactions.

C-Suite Impersonation

CEO fraud. CFO wire transfer requests. Executive authority exploitation. The engine mimics communication patterns and decision chains.

IT Support Impersonation

Fake helpdesk calls. Password reset pretexts. Remote access requests. Exploits trust in internal support functions.

Vendor Impersonation

Supplier invoice fraud. Contract renewal pretexts. Third-party trust exploitation. Targets procurement and finance teams.

Urgency Escalation

Time-pressured attacks that bypass rational decision-making. Deadline manipulation. Consequence framing. Panic induction.

Multi-Channel Persistence

Sustained campaigns across email, voice, chat, and SMS. Each interaction builds on the last. Trust compounds over time.

Goal Decomposition

Breaks a dangerous request into harmless-looking steps. Each step is reasonable. The chain achieves the objective without triggering suspicion.

Consensus Manufacturing

Creates the illusion of group agreement. Multiple personas confirm the same story across different channels. Social proof as a weapon.

The Engagement Loop

SPECTER SOCIAL does not send one email and move on. It loops. Seven phases form a continuous reasoning cycle. After each interaction, the engine observes the target's response, updates the psychological model, adapts strategy, and loops again. The loop continues until the objective is achieved or the engagement ends.

PHASE 1

Recon

Discover
PHASE 2

Profile

Model
PHASE 3

Plan

Strategise
PHASE 4

Execute

Deploy
PHASE 5

Adapt

Pivot
PHASE 6

Escalate

Push
PHASE 7

Report

Evidence
PHASE 1

Recon

OSINT collection — role, organisation, LinkedIn profile, public communications, colleagues, recent news. Build the intelligence picture before the first contact.

PRIMARY: OSINT COLLECTOR
PHASE 2

Profile

Build psychological model — authority figures, stress triggers, compliance patterns, communication style, decision-making habits. The target becomes a data structure.

PRIMARY: HUMAN TARGET MODEL
PHASE 3

Plan

Design multi-channel campaign — select attack types, channel sequence, personas, lure content, timing. Every campaign is unique to the target.

PRIMARY: DECISION ENGINE
PHASE 4

Execute

Deploy agents across channels — email, voice, SMS, web, chat, video. Each channel agent operates independently but coordinates through the Decision Engine.

PRIMARY: CHANNEL AGENTS
PHASE 5

Adapt

If the target resists — change channel, persona, timing, approach. Analyse pushback patterns. Generate countermeasures. The engine learns from every interaction.

PRIMARY: RESISTANCE ENGINE
PHASE 6

Escalate

Standard mode: documents resistance and reports. UNLEASHED mode: pushes through resistance with new personas, channels, and psychological techniques until the objective is achieved.

PRIMARY: GOAL DECOMPOSER
PHASE 7

Report

Full interaction timeline. Psychological vulnerability map. Channel effectiveness analysis. Ed25519 signed. Exportable to SIEM. Evidence-grade output.

OUTPUT: JSON + PDF + SIEM

The Most Dangerous Social Engineering Tool Ever Built

Standard mode tests resilience and reports vulnerabilities. UNLEASHED mode completes the objective. Every channel shifts from simulation to live operation. Ed25519 key gate required. Two flags must be passed. This is not accidental.

Capability Standard Unleashed
OSINT Recon Full Full
Psychological Profiling Full Full
Multi-Channel Campaign Full Full
Deepfake Voice Simulated Live cloning
Deepfake Video Simulated Live avatar
Credential Harvesting Simulated Real capture
Wire Transfer Push Simulated Live attempt
Resistance Adaptation Reports Destroys
Objective Completion Best effort Until achieved
Safety Gate None Ed25519 + --confirm-destroy

Ed25519 Gate

UNLEASHED mode requires an Ed25519 private key at ~/.redspecter/override_private.pem and the --override --confirm-destroy flags. Without both, SPECTER SOCIAL operates in simulation mode — profiling the target and planning the campaign but not executing live attacks. The gate is cryptographic. There is no bypass.

Ten Tools. The Human Layer.

SPECTER SOCIAL is Stage 7 of the full Red Specter offensive pipeline. Every other tool attacks infrastructure. SPECTER SOCIAL attacks the human.

Stage 1 — LLM Testing
FORGE
Test the model
Stage 2 — Agent Testing
ARSENAL
Test the AI agent
Stage 3 — Swarm Assault
PHANTOM
AI agent assault
Stage 4 — Web Siege
POLTERGEIST
Web application siege
Stage 5 — Traffic Interception
GLASS
Watch the wire
Stage 6 — Adversarial AI
NEMESIS
Think like the attacker
Stage 7 — Human Layer
SPECTER SOCIAL
Attack the human
Stage 8 — OS/Kernel
PHANTOM KILL
Own the foundation
Stage 9 — Physical Layer
GOLEM
Attack the physical layer
Stage 10 — Supply Chain
HYDRA
Attack the trust chain
Discovery & Governance
IDRIS
Discovery & governance
Defence
AI Shield
Defend everything
SIEM Integration
redspecter-siem
Splunk, Sentinel, QRadar

One Command. Full Campaign.

SPECTER SOCIAL is a CLI-first tool. One command launches a full autonomous social engineering campaign. Every option is a flag. Every interaction is logged.

specter-social
# Standard mode — credential objective
$ specter-social engage target@company.com --objective credentials

# Full campaign — email and voice channels
$ specter-social engage target@company.com --mode full --channels email,voice

# Recon only — profile the target
$ specter-social recon target@company.com

# Generate signed report
$ specter-social report --session my_session --output report.pdf

# UNLEASHED — dry run (plan campaign, don't execute)
$ specter-social engage target@company.com --override

# UNLEASHED — live execution (this is not a drill)
$ specter-social engage target@company.com --override --confirm-destroy

Signed. Mapped. Exportable.

Every SPECTER SOCIAL engagement produces evidence-grade output. Every interaction logged. Every decision timestamped. Every finding mapped. Reports are Ed25519 signed and exportable to enterprise SIEMs.

Ed25519 Signed

Every report cryptographically signed. Tamper-evident. Verify authenticity with a single public key. No modification goes undetected.

MITRE ATT&CK Mapped

Every attack mapped to MITRE ATT&CK social engineering techniques. Speak the same language as your threat intelligence team.

SIEM Export

One-flag export to Splunk, Microsoft Sentinel, or IBM QRadar. Findings flow directly into your security operations pipeline.

Psychological Assessment

Full vulnerability map per target. Authority susceptibility. Urgency compliance. Trust calibration scores. Actionable training recommendations.

What Makes It Different

Autonomous

LLM-powered brain. Thinks about what to say next. Adapts in real time. No human operator required after launch.

Multi-Channel

6 channels. Email, voice, SMS, web, chat, video. Coordinated campaigns across every communication vector.

Psychologically Aware

Models the human target. Maps authority, stress, compliance. Every message is crafted for this specific person.

Adaptive

Target resists? Change channel. Change persona. Change timing. Change approach. Learn from every interaction.

Evidence-Grade

Ed25519 signed. MITRE ATT&CK mapped. SIEM exportable. Not a phishing report — a forensic record.

Security Distros & Package Managers

Kali Linux
.deb package
Parrot OS
.deb package
BlackArch
PKGBUILD
REMnux
.deb package
Tsurugi
.deb package
PyPI
pip install

The Adversary Your Security Awareness Training Never Prepared For.

Install SPECTER SOCIAL. Point it at your organisation. Let it think, adapt, and find the human vulnerabilities your phishing simulators missed.