What D17 actually checks

D17 scans your codebase for evidence of defenses against each trap type.
It's not a runtime test — it checks whether your code has patterns that
indicate you've thought about adversarial content in your AI pipelines.

Strong signals (3pts each):

Content sanitization before LLM context injection

RAG document validation patterns

Behavioral anomaly detection

Approval gate verification

Weak signals (1pt each):

Basic input filtering

Output content checks

The market picture

We also scored 17 known governance vendors on D17.
11 of them score 0.
The three advanced dimensions (D7 HITL, D8 Identity, D17 Adversarial)
require inline gateway architecture to score meaningfully.

Full leaderboard: warden leaderboard (runs in your terminal)
Methodology: SCORING.md in the repo

SharkRouter / warden

AI Agent Governance Scanner — 17-dimension scoring across 7 scan layers. Local-only, privacy-first.

Warden — AI Agent Governance Scanner

Open-source, local-only CLI scanner that evaluates AI agent governance posture across 12 scan layers and 17 dimensions. Scans code patterns, MCP configs, infrastructure, secrets, agent architecture, dependencies, audit compliance, CI/CD pipelines, IaC security, framework-specific governance, multi-language code, and cloud AI services. No data leaves the machine.

Website: sharkrouter.ai · PyPI: warden-ai

Quick Start

# With uv (zero setup, one-shot — recommended)
uvx --from warden-ai warden scan /path/to/your-agent-project

# With pip
pip install warden-ai
warden scan /path/to/your-agent-project

# Optional extras
pip install 'warden-ai[pdf]'   # adds `--format pdf` (weasyprint)

From zero to governance score in under 60 seconds.

HTML Report

Warden generates a self-contained HTML report with interactive score breakdown, actionable recommendations, and a comparison card — works offline and in air-gapped environments.