naftiko: 1.0.0-alpha2
info:
  label: Lakera Guard — Content Screening
  description: 'Lakera Guard — Content Screening. 2 operations. Lead operation: Screen Content For Threats. Self-contained
    Naftiko capability covering the Lakera Guard runtime screening surface.'
  tags:
  - Lakera
  - Guard
  - AI Security
  - Prompt Injection
  created: '2026-05-25'
  modified: '2026-05-25'
binds:
- namespace: env
  keys:
    LAKERA_GUARD_API_KEY: LAKERA_GUARD_API_KEY
capability:
  consumes:
  - type: http
    namespace: guard-screening
    baseUri: https://api.lakera.ai/v2
    description: Lakera Guard content screening capability. Self-contained, no shared references.
    resources:
    - name: guard
      path: /guard
      operations:
      - name: screenContent
        method: POST
        description: Screen Content For Threats
        outputRawFormat: json
        outputParameters:
        - name: result
          type: object
          value: $.
        inputParameters:
        - name: body
          in: body
          type: object
          description: Request body (JSON) with messages, optional project_id, policy_id, breakdown, payload.
          required: true
    - name: guard-results
      path: /guard/results
      operations:
      - name: getResults
        method: POST
        description: Get Detector Confidence Results
        outputRawFormat: json
        outputParameters:
        - name: result
          type: object
          value: $.
        inputParameters:
        - name: body
          in: body
          type: object
          description: Request body (JSON) with messages and optional project_id.
          required: true
    authentication:
      type: apikey
      key: Authorization
      value: 'Bearer {{env.LAKERA_GUARD_API_KEY}}'
      placement: header
  exposes:
  - type: rest
    namespace: guard-screening-rest
    port: 8080
    description: REST adapter for Lakera Guard content screening. One Spectral-compliant resource per consumed
      operation, prefixed with /v2.
    resources:
    - path: /v2/guard
      name: guard
      description: REST surface for /guard.
      operations:
      - method: POST
        name: screenContent
        description: Screen Content For Threats
        call: guard-screening.screenContent
        with:
          body: rest.body
        outputParameters:
        - type: object
          mapping: $.
    - path: /v2/guard/results
      name: guard-results
      description: REST surface for /guard/results.
      operations:
      - method: POST
        name: getResults
        description: Get Detector Confidence Results
        call: guard-screening.getResults
        with:
          body: rest.body
        outputParameters:
        - type: object
          mapping: $.
  - type: mcp
    namespace: guard-screening-mcp
    port: 9090
    transport: http
    description: MCP adapter for Lakera Guard content screening. One tool per consumed operation.
    tools:
    - name: lakera-screen-content
      description: Screen LLM messages for prompt attacks, data leakage, content violations, PII, and malicious links.
      hints:
        readOnly: true
        destructive: false
        idempotent: true
      call: guard-screening.screenContent
      with:
        body: tools.body
      outputParameters:
      - type: object
        mapping: $.
    - name: lakera-get-detector-results
      description: Retrieve Lakera detector confidence levels for offline threshold tuning and quality monitoring.
      hints:
        readOnly: true
        destructive: false
        idempotent: true
      call: guard-screening.getResults
      with:
        body: tools.body
      outputParameters:
      - type: object
        mapping: $.