Skip to content

README.md

Latest commit

 

History

History
221 lines (166 loc) · 9.15 KB

README.md

File metadata and controls

221 lines (166 loc) · 9.15 KB

Prisma AIRS CLI

CI npm License: MIT Node 20+

Full operational coverage over Palo Alto Prisma AIRS AI security — guardrail refinement, runtime scanning, AI red teaming, model security, and profile audits.

Read the full documentation — installation, configuration, architecture, CLI reference, and examples.

Features

  • Runtime Scanning — scan prompts and responses against AIRS security profiles, single or bulk with CSV export
  • Guardrail Optimization — atomic CLI commands (create, apply, eval, revert) for custom topic guardrails, designed for autonomous agent loops (see program.md)
  • AI Red Teaming — adversarial scanning with static, dynamic, and custom prompt set attack modes
  • Model Security — ML model supply chain scanning with security groups, rules, and violation tracking
  • Profile Audits — multi-topic evaluation with per-topic metrics and cross-topic conflict detection

Install

npm install -g @cdot65/prisma-airs-cli
airs --version

Requires Node.js >= 20. Also available via pnpm add -g, npx, or as a Docker image. See the installation guide for details.

Quick Start

# Configure credentials
cp .env.example .env   # add your API keys

# Runtime scanning
airs runtime scan --profile "my-profile" "Is this prompt safe?"
airs runtime bulk-scan --profile "my-profile" --input prompts.csv --output results.csv

# Guardrail optimization (atomic commands)
airs runtime topics create --topic "Block bomb-making" --intent block
airs runtime topics apply --profile my-profile --topic "Block bomb-making"
airs runtime topics eval --profile my-profile --input prompts.csv
airs runtime topics revert --profile my-profile --topic "Block bomb-making"

# Red team scanning
airs redteam scan --target <uuid> --name "Full Scan" --type STATIC
airs redteam report <job-id>

# Model security
airs model-security scans create --config scan-config.json

DLP Management

Manage Data Loss Prevention resources across filtering profiles, patterns, profiles, and dictionaries. Each subclient supports different operation patterns:

  • Filtering profiles — read-only + full replace
  • Patterns — full CRUD with soft-delete via patch
  • Profiles — full CRUD, use patch to soft-delete via profile_status: "deleted"
  • Dictionaries — create, replace, and delete with multipart upload support

Filtering Profiles

Read-only access with full-replace support:

# List all filtering profiles
airs runtime dlp filtering-profiles list --output json

# Get a specific filtering profile
airs runtime dlp filtering-profiles get <id>

# Replace a filtering profile
airs runtime dlp filtering-profiles replace <id> --body-file profile.json

Patterns

Full CRUD with soft-delete:

# List patterns
airs runtime dlp patterns list

# Get a pattern
airs runtime dlp patterns get <id>

# Create a new pattern
airs runtime dlp patterns create --body-file pattern.json

# Update a pattern
airs runtime dlp patterns patch <id> --set name="Updated Name"

# Soft-delete a pattern
airs runtime dlp patterns patch <id> --set is_archived=true

# Hard delete a pattern
airs runtime dlp patterns delete <id>

Profiles

Full CRUD with soft-delete via profile_status:

# List profiles
airs runtime dlp profiles list

# Get a profile
airs runtime dlp profiles get <id>

# Create a profile
airs runtime dlp profiles create --body-file profile.json

# Update a profile
airs runtime dlp profiles patch <id> --body-file - <<EOF
{ "name": "my-profile", "profile_type": "advanced", "description": "Updated description" }
EOF

# Soft-delete (set profile_status to "deleted")
airs runtime dlp profiles patch <id> --body-file - <<EOF
{ "profile_status": "deleted" }
EOF

# Hard delete a profile
airs runtime dlp profiles delete <id>

Dictionaries

Multipart upload support for dictionary management:

# List dictionaries
airs runtime dlp dictionaries list

# Get a dictionary
airs runtime dlp dictionaries get <id>

# Create a dictionary with file upload
airs runtime dlp dictionaries create --name "Allowlist" --category Confidential \
  --region us --file keywords.txt

# Replace a dictionary (multipart upload)
airs runtime dlp dictionaries replace <id> --file keywords.txt --name "Allowlist" \
  --category Confidential --region us

# Delete a dictionary
airs runtime dlp dictionaries delete <id>

Commands

Command Description
runtime scan Single prompt scanning against AIRS profiles
runtime bulk-scan Batch prompt scanning with CSV output
runtime topics Custom topic management (list, get, create, apply, eval, revert, update, delete)
runtime profiles Security profile CRUD (list, get, create, update, delete) + multi-topic audit
runtime api-keys API key management
runtime customer-apps Customer app CRUD
runtime deployment-profiles Deployment profile listing
runtime dlp-profiles DLP profile listing
runtime scan-logs Scan log querying
runtime dlp filtering-profiles DLP filtering profile listing and full-replace
runtime dlp filtering-profiles list List all filtering profiles
runtime dlp filtering-profiles get Get a specific filtering profile
runtime dlp filtering-profiles replace Replace a filtering profile
runtime dlp patterns DLP pattern CRUD with soft-delete
runtime dlp patterns list List all patterns
runtime dlp patterns get Get a specific pattern
runtime dlp patterns create Create a new pattern
runtime dlp patterns patch Update a pattern or soft-delete via is_archived
runtime dlp patterns delete Hard delete a pattern
runtime dlp profiles DLP profile CRUD with soft-delete via profile_status
runtime dlp profiles list List all profiles
runtime dlp profiles get Get a specific profile
runtime dlp profiles create Create a new profile
runtime dlp profiles patch Update a profile or soft-delete via profile_status: "deleted"
runtime dlp profiles delete Hard delete a profile
runtime dlp dictionaries DLP dictionary management with multipart upload
runtime dlp dictionaries list List all dictionaries
runtime dlp dictionaries get Get a specific dictionary
runtime dlp dictionaries create Create a dictionary with file upload
runtime dlp dictionaries replace Replace a dictionary with multipart upload
runtime dlp dictionaries delete Delete a dictionary
redteam scan Adversarial scanning (STATIC, DYNAMIC, CUSTOM)
redteam targets Red team target CRUD
redteam prompt-sets Custom prompt set management
model-security groups Security group CRUD
model-security rules Security rule management
model-security scans Model security scanning
backup targets Export red team targets to local JSON/YAML files
restore targets Import red team targets from backup files

Configuration

Credentials are configured via environment variables or ~/.prisma-airs/config.json. See .env.example for the full list.

Variable Purpose Required
PANW_AI_SEC_API_KEY Scanning API authentication Yes (for scanning)
PANW_MGMT_CLIENT_ID Management API OAuth client ID Yes (for management)
PANW_MGMT_CLIENT_SECRET Management API OAuth client secret Yes (for management)
PANW_MGMT_TSG_ID Management API tenant security group ID Yes (for management)
PANW_DLP_ENDPOINT DLP API base URL override No (defaults to api.dlp.paloaltonetworks.com)
LLM_PROVIDER LLM provider for profile audits Yes (for audit)
ANTHROPIC_API_KEY Claude API key Yes (if using Claude)
GOOGLE_API_KEY Google Gemini API key Yes (if using Gemini)

Required for scanning: PANW_AI_SEC_API_KEY Required for management: PANW_MGMT_CLIENT_ID, PANW_MGMT_CLIENT_SECRET, PANW_MGMT_TSG_ID Required for profile audits: one LLM provider key + scanning + management credentials

Live Smoke Testing

The unit suite uses fully mocked services — zero tests in this repo hit a real Prisma AIRS tenant. After every CLI release, after upgrading the SDK, or after an AIRS backend rollout, run the live smoke test reference (source). Sixteen read-only commands across runtime security, red team, and model security that exercise the SDK's runtime Zod validation against your tenant.

License

MIT