FakeLiang

FakeLiang is an attack graph validation agent that tests paths on ranges using multi-agent collaboration.

FakeLiang reads an attack graph, asset information, and a host-IP mapping table to verify graph paths on a target range. It uses skill cards to drive a large language model through planning, execution, and verification across multiple agents. The system saves call parameters, standard output, error output, and model transcriptions for every step. It runs on Windows with Python, requiring a Kali instance with SSH and Metasploit RPC, plus an OpenAI-compatible model endpoint. The product outputs JSON, DOT, and PDF files detailing node statuses and run summaries.

FakeLiang at a glance

Best for: Software engineer
Runs on: Windows, Python, CLI
Pricing: Free
Released: October 2026
Category: AI agent tool
Website: github.com/cxm947/fakeliang

Key features

Drive path progression from attack graph nodes and dependencies
Use skill cards containing conditions, inputs, and actions
Collaborate across Planner, Executor, and Verifier agents
Save call parameters, standard output, and model transcriptions
Output JSON, DOT, and PDF graph visualizations
Run offline checks for tests, skill cards, and file maps

Use it for

Validate attack paths on target ranges automatically
Run single nodes or individual skill cards
Manage sessions and reclaim resources
Read logs and model transcriptions

Alternatives to FakeLiang

iFixAi — Audits agent behavior independently without building manual eval harnesses.
ARTEX — Runs autonomous multi-agent penetration testing with a dual-graph architecture.
TaintWall — Enforces deterministic security policies at the tool boundary for AI agents.
Truth Firewall — Blocks unverified coding agent completion claims automatically.
ARTEX — Manages autonomous penetration testing tasks and tracks vulnerability progress.
Sapper V2 — Beats traditional agents with parallel roles and layered fact graphs.
Autonomous Pentest Agent — Dual-graph architecture for asset synchronization and multi-step attack planning.
CyberStrike — Autonomous AI agents execute recon, web, network, and cloud penetration testing.
AgentTrust — Enforces zero-trust policy rules and maintains an immutable cryptographic audit ledger.
Security Audit Skill — Automated multi-stage security auditing with independent verification.
Artex Global — Autonomous multi-agent penetration testing with multilingual output support.
TrustGraph — Provides ontology-grounded knowledge for auditable and compliant AI workloads.

All FakeLiang alternatives