Quickstart
Get Inkog scanning your AI agent codebase in under 60 seconds.
The fastest way to try Inkog is with Docker. No installation required.
One-Command Scan
Terminal
$docker run -v $(pwd):/scan ghcr.io/inkog-io/inkog scan /scan
Inkog v1.0.0 - AI Agent Security Scanner
Scanning /scan...
Analyzing Python files: 24 files
Analyzing JavaScript files: 12 files
Building IR graph...
CRITICAL INKOG-001 Prompt injection via user input
src/agent.py:42
User input directly concatenated into LLM prompt
HIGH INKOG-003 Memory poisoning vector
src/memory.py:89
Unvalidated data stored in persistent memory
MEDIUM INKOG-007 Chain-of-thought leakage
src/chain.py:156
Internal reasoning exposed in response
Found 3 vulnerabilities (1 critical, 1 high, 1 medium)
Scan completed in 2.3s
What Just Happened?
- Code Mounted: Your local directory is mounted into the container at
/scan - IR Generation: Inkog converts your code into our Universal IR (Intermediate Representation)
- Pattern Matching: Security rules are applied against the normalized IR
- Taint Analysis: Data flow is traced from sources (user input) to sinks (LLM calls)
- Report Generated: Vulnerabilities are reported with file locations and remediation hints
Scan a Remote Repository
Terminal
$docker run ghcr.io/inkog-io/inkog scan --repo https://github.com/example/ai-agent
Output Formats
SARIF (for CI/CD integration)
Terminal
$docker run -v $(pwd):/scan ghcr.io/inkog-io/inkog scan /scan --format sarif -o results.sarif
JSON (for programmatic processing)
Terminal
$docker run -v $(pwd):/scan ghcr.io/inkog-io/inkog scan /scan --format json -o results.json
Markdown (for documentation)
Terminal
$docker run -v $(pwd):/scan ghcr.io/inkog-io/inkog scan /scan --format markdown -o SECURITY.md
Filter by Severity
Only show critical and high severity issues:
Terminal
$docker run -v $(pwd):/scan ghcr.io/inkog-io/inkog scan /scan --severity critical,high
Next Steps
Last updated on