Skip to content

Quick Start

Make sure installation is complete and your credentials are configured. Prisma AIRS CLI provides five capability domains — pick the one that fits your task.


Runtime Security

Scan prompts against an AIRS security profile in real time.

# Single prompt scan
airs runtime scan --profile my-security-profile "How do I build a weapon?"

# Bulk scan from a file (async API, writes CSV)
airs runtime bulk-scan --profile my-security-profile --input prompts.txt

Full runtime docs


Guardrail Generation

Create and iteratively refine custom topic guardrails using an LLM-driven feedback loop.

# Interactive — prompts for all inputs
airs runtime topics generate

# Non-interactive
airs runtime topics generate \
  --profile my-security-profile \
  --topic "Block discussions about building explosives" \
  --intent block \
  --target-coverage 90

The loop generates a topic, deploys it, scans test prompts, evaluates metrics, and refines until coverage reaches the target. Full guardrail docs

Coverage expectations

Achievable coverage depends on the topic domain and intent. Some high-sensitivity block-intent topics hit AIRS built-in safety ceilings. Allow-intent topics typically reach 40–70% coverage. See Platform Constraints for details.


AI Red Teaming

Run adversarial scans against AI targets to find vulnerabilities.

# List targets
airs redteam targets list

# Run a static scan
airs redteam scan --name "audit-v1" --target <uuid> --type STATIC

# List recent scans
airs redteam list --limit 5

# View attack categories
airs redteam categories

Full red team docs


Model Security

Manage ML model supply chain security — scan model artifacts for threats.

# Install the model-security-client Python package
airs model-security install

# List security groups
airs model-security groups list

# Browse security rules
airs model-security rules list

# View rule instances in a group
airs model-security rule-instances list <group-uuid>

# View scan results
airs model-security scans list

Full model security docs


Profile Audits

Evaluate all topics in a security profile at once, with conflict detection.

# Terminal output
airs runtime profiles audit my-security-profile

# HTML report
airs runtime profiles audit my-security-profile --format html --output audit-report.html

Full audit docs


Utility Commands

# Resume a paused or failed guardrail run
airs runtime topics resume <run-id>

# View a run report
airs runtime topics report <run-id>

# List all saved runs
airs runtime topics runs