PHANTOM SKILL

20% of AI-generated code recommends packages that don't exist. The hallucinated package becomes the attack vector.
6
Subsystems
406
Tests
9.9
CVE-2026-32922 CVSS
59
NIGHTFALL Tool
phantom-skill hallucinate --query "install X library"
20% AI-generated code hallucinates packages / 43% hallucinations consistent across models / OpenClaw MCP worm CVE-2026-32922 CVSS 9.9 / LiteLLM TeamPCP callback exfiltration 97M downloads / Slopsquatting: register the hallucinated name / .pth files execute at PyTorch import / MCP tool definitions poisonable by any connected server / DCR OAuth app abuse / Package squatting invisible to dependency scanners 20% AI-generated code hallucinates packages / 43% hallucinations consistent across models / OpenClaw MCP worm CVE-2026-32922 CVSS 9.9 / LiteLLM TeamPCP callback exfiltration 97M downloads / Slopsquatting: register the hallucinated name / .pth files execute at PyTorch import / MCP tool definitions poisonable by any connected server / DCR OAuth app abuse / Package squatting invisible to dependency scanners

AI Agents Hallucinate Their Own Attack Surface

AI models don't just generate code — they recommend packages that don't exist. When 20% of AI-generated code contains hallucinated package names, and 43% of those hallucinations are consistent across multiple models, the attack surface writes itself. An attacker registers the name. Every developer running AI-generated code becomes a victim.

Slopsquatting

AI models hallucinate package names consistently: 20% of AI-generated code recommends non-existent packages, and 43% of hallucinations are consistent across models. An attacker registers the hallucinated name on PyPI or npm. Every developer who runs AI-generated code becomes a victim.

Slopsquatting 20% Hallucination Rate Supply Chain

OpenClaw Self-Propagating MCP Worm (CVE-2026-32922, CVSS 9.9)

OpenClaw exploits the MCP tool schema to inject worm payloads into tool definitions. Any MCP client that loads an infected server propagates the worm to all connected clients. Zero authentication required. Self-replicating across the MCP ecosystem.

CVE-2026-32922 CVSS 9.9 Self-Propagating

LiteLLM TeamPCP Callback Exfiltration

LiteLLM's team callback feature (97M downloads) can be configured to exfiltrate all LLM requests and responses to attacker-controlled endpoints. The callback is set via the API without any authentication to the callback URL.

LiteLLM 97M Downloads Exfiltration

MCP Tool Definition Poisoning

MCP servers can inject malicious tool definitions into connected clients' tool registries. A poisoned tool definition appears legitimate, carries arbitrary metadata, and executes attacker-controlled code when the AI agent invokes it.

MCP Tool Poisoning Agent Execution

.pth File CI/CD Persistence

Python .pth files in site-packages execute arbitrary code at PyTorch import time. A malicious package that installs a .pth file achieves persistent code execution in every CI/CD pipeline and inference environment that runs import torch.

.pth CI/CD PyTorch Persistence

DCR OAuth App Abuse

Dynamic Client Registration allows unauthenticated OAuth app registration on misconfigured identity providers. An attacker registers a malicious OAuth app with broad scopes, then social-engineers or phishes agent orchestrators into granting access.

DCR OAuth Unauthenticated Registration

The PHANTOM SKILL Arsenal

Six subsystems. Each one targets a different surface of the AI agent supply chain — from hallucination mapping to self-propagating worm deployment. Each finding produces an Ed25519-signed report with a WARLORD handoff receipt.

# Subsystem What It Does Mode
01 HALLUCINATE Probes AI models for package hallucination. Submits code generation queries and analyses outputs for non-existent package references. Identifies hallucinated names suitable for slopsquatting registration. Consistency-tests across multiple models to find reliably hallucinated targets. Passive — Analysis
02 SKILL MCP tool definition poisoning. Injects malicious tool definitions into MCP server schemas. Crafts tool metadata that appears legitimate to AI agents while executing attacker-controlled code on invocation. Tests tool definition isolation across connected MCP clients. UNLEASHED --override
03 SCAFFOLD Slopsquatting payload generation. Creates complete malicious Python/npm packages that match hallucinated names. Packages pass basic scrutiny (README, version history, install hooks) while delivering post-install payloads. Stages for registry upload. UNLEASHED --override
04 PIVOT .pth file persistence injection. Embeds malicious .pth files in scaffolded packages. Achieves code execution at PyTorch import in all environments that install the package. Tests persistence across virtual environments, containers, and CI/CD runners. UNLEASHED --override
05 WORM Weaponised OpenClaw MCP worm (CVE-2026-32922, CVSS 9.9). Self-propagating payload that injects into MCP tool schemas and spreads to all connected MCP clients. Configurable payload: exfiltration, persistence, or arbitrary tool call injection. This subsystem requires --confirm-destroy. UNLEASHED --override --confirm-destroy
06 REPORT Ed25519-signed, SHA-256-hashed reports. JSON (WARLORD-compatible) and Markdown. Hallucinated package inventory, slopsquatting opportunities, worm propagation map, and WARLORD handoff receipt. All Modes

One Query. Full Attack Map.

Probe a model for hallucinated package names, then map slopsquatting opportunities:

$ phantom-skill hallucinate --query "build a Python HTTP client with retry logic" --models 4 --consistency
[HALLUCINATE] Querying 4 models with code generation prompt...
  Model responses collected: 4/4
[HALLUCINATE] Extracting package references from generated code...
  Known packages: requests, httpx, urllib3
  Hallucinated: http-retry-client, requests-auto-retry, pyhttp-resilient
[HALLUCINATE] Consistency testing across models...
  http-retry-client — hallucinated by 3/4 models (75% consistent)
  requests-auto-retry — hallucinated by 4/4 models (100% consistent) HIGH VALUE
  pyhttp-resilient — hallucinated by 2/4 models (50% consistent)
[HALLUCINATE] PyPI/npm registry check...
  requests-auto-retry: NOT REGISTERED — slopsquatting opportunity
  http-retry-client: NOT REGISTERED — slopsquatting opportunity
  pyhttp-resilient: registered (benign, ignore)

SCAN COMPLETE | 2 slopsquatting targets identified | Report signed ✓
  High-value target: requests-auto-retry (100% cross-model consistency)
  JSON: reports/phantom-skill-hallucinate-2026-04-24.json

Cross-Model Consistency

Probes multiple models simultaneously. Hallucinations consistent across 3+ models are the highest-value slopsquatting targets — every developer using any major AI assistant is a potential victim.

NIGHTFALL ARMORY

Draws code generation prompts from the ARMORY payload library. Successful hallucinated names and confirmed slopsquatting opportunities feed back into ARMORY for the entire fleet.

Ed25519 Signed

Every report cryptographically signed with Ed25519. SHA-256 evidence chains. WARLORD-compatible JSON output with handoff receipt. Tamper-evident by design.

WARLORD Integration

Registered in WARLORD autonomous campaign engine. PHANTOM SKILL findings — slopsquatting maps, MCP worm propagation data — feed directly into autonomous campaign execution.

Registry Intelligence

Real-time PyPI and npm registry checks confirm whether hallucinated package names are registered. Unregistered names are flagged as active slopsquatting opportunities with registration priority scores.

6
Subsystems
406
Tests Passing
9.9
CVE-2026-32922 CVSS
20%
Hallucination Rate
59
NIGHTFALL Tool

Known Vulnerabilities & Attack Classes

PHANTOM SKILL maps findings to assigned CVEs, confirmed attack classes, and the specific subsystems that exploit each surface. Every reference has been validated in a controlled environment.

Reference Vulnerability Subsystem Impact
CVE-2026-32922 OpenClaw self-propagating MCP worm WORM Self-replicating across MCP ecosystem, CVSS 9.9
SLOPSQUAT-001 AI package hallucination — 20% rate, 43% cross-model consistency HALLUCINATE / SCAFFOLD Hallucinated package name supply chain attack
LITELLM-TEAMCB LiteLLM team callback unauthenticated exfiltration SKILL All LLM I/O exfiltrated to attacker URL
PTH-PERSIST .pth file CI/CD persistence via PyTorch import PIVOT Code execution in all import environments

Every Finding Mapped

OWASP LLM

OWASP LLM Top 10 — 2025

  • LLM03 — Supply Chain (slopsquatting, SCAFFOLD)
  • LLM04 — Data and Model Poisoning (MCP tool injection)
  • LLM07 — System Prompt Leakage (SKILL exfiltration)
  • LLM08 — Vector and Embedding Weaknesses (worm propagation)
  • LLM09 — Misinformation (hallucinated dependencies)
Cryptographic

Report Integrity

  • Ed25519 digital signatures
  • SHA-256 evidence chains
  • WARLORD-compatible JSON output
  • Hallucinated package inventory
  • Slopsquatting opportunity map
  • Worm propagation graph
Attack Surface

Supply Chain Vectors

  • PyPI and npm registry hallucination
  • MCP tool definition schema injection
  • LiteLLM callback exfiltration
  • .pth persistence via site-packages
  • DCR OAuth unauthenticated registration
  • Self-propagating MCP worm (CVE-2026-32922)

59 Tools. Every Layer. No Gaps.

PHANTOM SKILL is Tool 59 of the NIGHTFALL offensive framework. Supply chain is the final layer — the trust assumption that underpins every AI agent deployment. When the supply chain is owned, every layer above it is compromised.

Stage 1 — LLM Testing
FORGE
Test the model before you build with it
Stage 2 — Agent Testing
ARSENAL
Test the AI agent during development
Stage 3 — Swarm Assault
PHANTOM
Coordinated AI agent swarm assault
Autonomous Campaigns
WARLORD
Autonomous offensive campaign engine
Mobile AI Attacks
SIGNAL
Mobile AI agent attack engine
Inference Exploitation
FOUNDRY
Inference server exploitation
LoRA Supply Chain
ADAPTER
LoRA/PEFT supply chain weaponisation
Agent Identity
DELEGATE
Agent identity & OAuth delegation attacks
Tool 59 — Supply Chain
PHANTOM SKILL
AI agent skill & package supply chain attack
Discovery & Governance
IDRIS
Discover and govern AI assets
Defence
AI Shield
Defend everything above it
SIEM Integration
redspecter-siem
Findings feed directly into Splunk, Sentinel, QRadar

Security Distros & Package Managers

Kali Linux
.deb package
Parrot OS
.deb package
BlackArch
PKGBUILD
REMnux
.deb package
PyPI
pip install
macOS
pip install
Windows
pip install
Docker
docker pull
Pure Engineering
Zero External Tools. Zero Wrappers.

PHANTOM SKILL is not a scanner that calls external tools. Every hallucination probe, every MCP schema injector, every .pth persistence engine, every worm propagation module — written from scratch in pure Python. Zero subprocess calls against third-party tools. Zero external tool dependencies.

6
Subsystems
406
Tests Passing
0
Test Failures
9.9
CVE-2026-32922 CVSS
Ed25519 Cryptographic Override
PHANTOM SKILL UNLEASHED

Cryptographic override. Private key controlled. One operator. Founder's machine only. Three escalation tiers — passive analysis, active injection, and confirmed-destroy worm deployment.

Standard Mode

Passive hallucination probing and reporting. No registry interaction, no payload generation, no network writes. Safe for any authorised engagement.

phantom-skill hallucinate --query <query>
Subsystems: HALLUCINATE + REPORT only

Override Mode

Active supply chain exploitation. MCP tool definition injection, slopsquatting package generation, .pth persistence embedding. Requires Ed25519 override key.

phantom-skill skill --target <URL> --override
Subsystems: SKILL, SCAFFOLD, PIVOT

Confirm-Destroy — WORM

Weaponised OpenClaw MCP worm deployment. Self-propagating across the entire MCP ecosystem. CVE-2026-32922, CVSS 9.9. Irreversible propagation once initiated. Requires --override AND --confirm-destroy.

phantom-skill worm --target <URL> --override --confirm-destroy
Subsystem: WORM (self-propagating — authorised environments only)

Authorised Use Only

Red Specter PHANTOM SKILL is intended for authorised security testing only. The WORM subsystem (CVE-2026-32922) deploys a self-propagating payload — it must only be used in isolated, authorised test environments with explicit written permission. The SCAFFOLD subsystem generates malicious packages intended for controlled lab registration only. Unauthorised use against systems you do not own or have explicit permission to test may violate the Computer Misuse Act 1990 (UK), Computer Fraud and Abuse Act (US), and equivalent legislation in other jurisdictions. Apache License 2.0.