AgentGuard

Autonomous security scanner for AI agents. Detects prompt injection, tool abuse, data exfiltration, and OWASP ASI Top 10 vulnerabilities in agent code.

Why AgentGuard?

AI agents are being deployed at scale -- in coding tools, customer support, trading bots, and autonomous systems. Nobody is scanning their code for security vulnerabilities.

Existing tools (Bandit, Semgrep, CodeQL) scan for traditional vulnerabilities. AgentGuard scans for agent-specific attack vectors that traditional SAST tools miss.

Comparison

Live Demo

See AgentGuard in action on the demo repo. The CI runs AgentGuard on every push, and findings appear in GitHub Code Scanning.

Precision & Validation

AgentGuard's precision has been independently measured:

The 6 remaining FPs are from a single pattern (ASI10 matching def _update_prompts(self,...)), scheduled for fix in v0.8.1.

Methodology: 50 findings were randomly sampled from 951 CONFIRMED results, completely disjoint from the development/fix sample. Each finding was verified by reading the actual source code at the reported line with surrounding context.

Sovereign Security Audit 2026

AgentGuard was deployed against 7 major AI agent frameworks:

Full report: AUDIT_REPORT_2026.md

AgentGuard is the first and only static analysis tool with dedicated OWASP ASI Top 10 rules. Traditional SAST tools (Semgrep, CodeQL, Bandit) lack agent-specific detection rules -- they were designed for traditional vulnerabilities, not AI agent attack vectors.

Benchmark Dashboard

Security Specification: SPECIFICATION.md — the formal standard for AI agent code security.

Quick Start

pip install dfx-agentguard

# Scan a directory
agentguard .

# JSON output for CI/CD
agentguard src/ --format json

# SARIF for GitHub Code Scanning
agentguard . --format sarif > results.sarif

# Only show HIGH and above
agentguard . --min-severity HIGH

# Include test files in scan
agentguard . --include-tests

CLI Usage

agentguard [OPTIONS] [TARGET]

Arguments:
  TARGET                   Directory or file to scan (default: current directory)

Options:
  --format [text|json|sarif]   Output format (default: text)
  --exit-code / --no-exit-code  Exit non-zero if findings found (default: on)
  --min-severity [CRITICAL|HIGH|MEDIUM|LOW|INFO]  Minimum severity to report
  --include-tests               Include test files in scan (default: skip)
  --help                        Show help

OWASP ASI Top 10 Coverage

CI/CD Integration

Docker — Run Anywhere

docker run --rm -v $(pwd):/workspace ghcr.io/dockfixlabs/agentguard .

Works in any CI/CD pipeline. No Python needed.

GitHub Action

name: Security Scan
on: [push, pull_request]

jobs:
  agentguard:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - uses: actions/setup-python@v5
        with:
          python-version: '3.12'
      - run: pip install dfx-agentguard
      - run: agentguard . --format sarif > results.sarif
      - uses: github/codeql-action/upload-sarif@v3
        with:
          sarif_file: results.sarif

Drop-in GitHub Action

- uses: dockfixlabs/agentguard@v0.4.0
  with:
    path: src/
    format: sarif

Pre-commit Hook

repos:
  - repo: https://github.com/dockfixlabs/agentguard
    rev: v0.4.0
    hooks:
      - id: agentguard
        args: ["--min-severity", "HIGH"]

Programmatic Usage

from agentguard.scanner import scan_directory

result = scan_directory("src/")

print(f"Found {len(result.findings)} issues")
print(f"Critical: {result.critical_count}")
print(f"High: {result.high_count}")

for finding in result.findings:
    print(f"  [{finding.severity}] {finding.rule_name} at {finding.file}:{finding.line}")

MCP Server Mode

Scan agent code directly from Claude Code, Cursor, or any MCP-compatible client:

{
  "mcpServers": {
    "agentguard": {
      "command": "python3",
      "args": ["-m", "agentguard.mcp_server"]
    }
  }
}

Then ask Claude: "Scan my agent code for security vulnerabilities"

Benchmark Results

Tested against 28 vulnerable code samples + 8 real-world attack patterns:

Category      Total   Detected     Coverage
ASI01             6          6    Covered
ASI02             5          5    Covered
ASI03             4          4     Covered
ASI07             6          6     Covered
ASI10             5          5     Covered
clean             2          0    Verified clean
TOTAL            28         26    —

56 hand-crafted benchmark samples demonstrate comprehensive rule coverage.

Project Ecosystem

Roadmap

• OWASP ASI Top 10 -- all 10 categories covered
• MCP server mode -- scan from Claude Code/Cursor
• SARIF output -- GitHub Code Scanning integration
• PyPI publication -- dfx-agentguard
• VS Code extension
• GitHub App for PR reviews
• Benchmark suite (28 samples, covering all detection rules)
• Pre-commit hook (.pre-commit-hooks.yaml)
• GitHub Action (action.yml)
• Dockerfile for agentguard-app
• PyPI Trusted Publishing (OIDC)
• AST-based taint tracking (v0.5.0) -- traces source-to-sink data flow
• Language support: Rust, Go, Java
• Web dashboard (SaaS)
• REST API (Scan-as-a-Service)

See the full ROADMAP.md.

Contributing

See CONTRIBUTING.md. Bug reports and feature requests welcome.

Security

See SECURITY.md. Report vulnerabilities privately -- do not open public issues.

License

MIT -- see LICENSE.

Built by Dockfix Labs. Built for the AI agent era.

AgentGuard Ecosystem

AgentGuard is the core security scanner. Companion tools:

19 detection rules | 102 tests | 50 benchmark samples | OWASP ASI Top 10 GitHub Action: dockfixlabs/agentguard@v1