Configuration Reference¶

Configuration Hierarchy¶

1. Strata Cloud Manager (SCM)
   └── Security profiles, detection rules, actions, DLP patterns

2. OpenClaw Plugin Config
   └── api_key (required)
   └── profile_name, app_name
   └── Hook toggles, local enforcement behavior

Guardrails Are in SCM

This plugin does NOT configure AI guardrails. All detection services, sensitivity levels, and actions are configured in Strata Cloud Manager. The plugin simply points to your SCM security profile and enforces the actions it returns.

Plugin Configuration¶

Add to OpenClaw config (via gateway web UI or config file):

plugins:
  prisma-airs:
    enabled: true
    config:
      # API key (required)
      api_key: "your-api-key-here"

      # Core settings
      profile_name: "default"
      app_name: "openclaw"

      # Hook toggles
      reminder_enabled: true
      audit_enabled: true
      context_injection_enabled: true
      outbound_scanning_enabled: true
      tool_gating_enabled: true

      # Local enforcement
      fail_closed: true
      dlp_mask_only: true
      high_risk_tools:
        - exec
        - Bash
        - bash
        - write
        - Write
        - edit
        - Edit
        - gateway
        - message
        - cron

Core Settings¶

profile_name¶

Property	Value
Type	`string`
Default	`"default"`

Security profile name from Strata Cloud Manager.

app_name¶

Property	Value
Type	`string`
Default	`"openclaw"`

Application identifier included in scan metadata for SCM reporting.

Hook Toggles¶

Option	Type	Default	Hook
`reminder_enabled`	boolean	`true`	prisma-airs-guard
`audit_enabled`	boolean	`true`	prisma-airs-audit
`context_injection_enabled`	boolean	`true`	prisma-airs-context
`outbound_scanning_enabled`	boolean	`true`	prisma-airs-outbound
`tool_gating_enabled`	boolean	`true`	prisma-airs-tools

Local Enforcement Settings¶

These control how the plugin responds locally—NOT what AIRS detects.

fail_closed¶

Property	Value
Type	`boolean`
Default	`true`

When true, scan failures (API errors, timeouts) result in blocked requests. When false, failures allow requests through.

dlp_mask_only¶

Property	Value
Type	`boolean`
Default	`true`

When true, DLP violations are masked instead of blocked. When false, DLP violations block the response entirely.

Always-Block Categories

Regardless of dlp_mask_only, these categories always block: malicious_code*, malicious_url, toxicity, toxic_content*, agent_threat*, prompt_injection, db_security*, scan-failure (includes suffixed variants like malicious_code_response)

high_risk_tools¶

Property	Value
Type	`string[]`
Default	See below

Tools to block when ANY threat is detected (not just specific categories).

Default:

high_risk_tools:
  - exec
  - Bash
  - bash
  - write
  - Write
  - edit
  - Edit
  - gateway
  - message
  - cron

Strata Cloud Manager Settings¶

These are configured in SCM, not the plugin:

Setting	SCM Location	Description
Detection services	Security Profiles	Which threats to detect
Actions	Security Profiles	allow/alert/block per detection
Sensitivity	Security Profiles	Detection thresholds
DLP patterns	Data Loss Prevention	PII, credential patterns
URL categories	URL Filtering	Allowed/blocked URL categories
Custom topics	Topic Guardrails	Organization policies

Example: Minimal Setup¶

plugins:
  prisma-airs:
    config:
      api_key: "your-key"

All other settings use sensible defaults.

Example: Audit Only Mode¶

plugins:
  prisma-airs:
    config:
      audit_enabled: true
      context_injection_enabled: false
      outbound_scanning_enabled: false
      tool_gating_enabled: false
      fail_closed: false

Example: Maximum Local Enforcement¶

plugins:
  prisma-airs:
    config:
      fail_closed: true
      dlp_mask_only: false
      high_risk_tools:
        - exec
        - Bash
        - write
        - edit
        - gateway
        - message
        - cron
        - browser
        - WebFetch
        - eval
        - NotebookEdit