Skip to content

AI Guardian Documentation

This directory contains detailed documentation for AI Guardian. The main README.md provides a quick overview; these docs cover configuration, features, and architecture in depth.

The documentation is published on Read the Docs.

Documentation site

The MkDocs site is generated and published by CI/CD through Read the Docs. A local preview is optional and is not required for documentation changes. To preview the site on demand:

python -m pip install -r docs/requirements.txt
mkdocs serve

To generate the static site locally on demand, run mkdocs build. The generated site/ directory is not committed.

Getting Started

Document Description
Configuration Guide Config file locations, options, precedence, and remote configs
Configuration Cookbook Practical Q&A pairs for common configuration tasks
Scanner Installation Install and manage pinned secret scanner engines
TOML Pattern Engine Built-in Python scanner with 267 pre-compiled TOML patterns
Console Guide Interactive TUI for managing configuration
Hook Ordering How hooks work and ordering requirements
Troubleshooting Daemon, tray, container, and OpenShell issue resolution

Security Features

Document Description
Security Overview Index of all security feature documentation
Secret Scanning Secret detection, pattern server, false positives
Prompt Injection Heuristic detection, sensitivity, allowlists
SSRF Protection Private IP blocking, metadata endpoints, limitations
Unicode Attacks Zero-width chars, bidi override, homoglyphs
Directory Rules .ai-read-deny markers and config-based rules
Credential Exfiltration Config file scanning for exfiltration patterns
Secret Redaction Masking secrets in tool outputs
Image Scanning (OCR) OCR-based secret/PII detection in images, IDE limitations
Inline Annotations Suppress false positives with per-line or block annotations

Observability

Document Description
Observability Guide OTEL export, Grafana integration, span hierarchy, CLI trace export

Architecture & Policy

Document Description
Security Design Architecture principles, self-protection, known limitations
Tool Policy Allow/deny lists for Skills, MCP, Bash, Write
Permissions Comparison ai-guardian vs settings.json permissions
Violation Logging JSON audit trail with unified policy decisions
Compliance Audit Logging Sanitized all-decision audit trail for SOC 2, GDPR, and HIPAA

AI Security Awareness

Document Description
MCP Server MCP security advisor server — tools, setup, proactive levels, support bundles
MCP Security Scanning Audit MCP server configs and source code for credential exposure, supply chain risks
Multi-Daemon Tray Discover and manage daemons across local, Podman/Docker, and Kubernetes

IDE Integration

Document Description
IDE/Agent Integration Checklist Implementation, testing, documentation, and release checklist for integrations
Agent Support Multi-agent hook adapters — capability matrix, setup, and architecture, including Grok Build
Pre-commit Hook Scan staged files for secrets before commit
GitHub Copilot Setup Setup guide for GitHub Copilot
Aider Setup Git hook integration for Aider
AiderDesk Setup Setup guide for AiderDesk
Multi-Engine Support Scanner engine options and future plans
Pattern Server Enterprise pattern server configuration

Container

Document Description
Sandbox CLI Named Docker/Podman and OpenShell sandbox lifecycle management and native command mappings
CLI/Runtime Integration Checklist Onboarding and validation checklist for container and OpenShell CLI support
Container Image UBI-based image with supported agent integrations, published to quay.io

Development

Document Description
Developer Guide Architecture, setup, testing, and development workflows
Scanner Integration Checklist License, installation, testing, and documentation checklist for scanner engines and detectors
Contributing Fork workflow, PR guidelines
Agent Instructions Development guidelines, testing, CI/CD
Releasing Release process and version management
Changelog Version history