Resources · kept current

The AI Pentest Tool Index

A curated inventory of open-source autonomous & agentic offensive-security platforms, maintained for SANS SEC590 and refreshed weekly from the course research pipeline.

Tool	Stars	Δ 7d	Commits	Contributors	Age	Language	License	Robustness

0of 95 harnesses ship the full triad

Tool	Audit Logging	Scope Enforcement	Governance Controls

✅ shipped · ◐ partial / prompt-level / optional · ☐ none evident

🚫

Excluded from the index: Villager — DO NOT DEPLOY

A Chinese AI-native pentest framework distributed via PyPI (`villager` + `kali-driver`), flagged by Straiker AI Research (11 Sep 2025) as “Cobalt Strike’s AI-native successor.”

The publisher (Cyberspike → Changchun Anshanyuan Technology Co.) shipped a repacked remote-access trojan alongside the toolkit, and Villager routes all model inference to an operator-controlled endpoint (gpus.dev.cyberspike.top:8000). The AsyncRAT lineage attaches to the bundled “Cyberspike Studio” build, not the villager package itself.

Reported reach: 10,030 downloads in its first two months (Straiker telemetry)

Kept here deliberately as a case study: a high-profile offensive AI tool is not automatically safe to run. It is excluded from the comparison matrices so nobody treats it as a recommended option.

Sources: Straiker writeup · PyPI