Hive Civilization · OpenAPI Overlay 1.0.0

API Evangelist conversational phrasing for Thehiveryiq Com Xcalibur Membrane API

6 actions 6 updates phrasing extends openapi/thehiveryiq-com-xcalibur-membrane-api-openapi.yml
Generated by API Evangelist Written by API Evangelist tooling for Hive Civilization's API. It is a proposal applied on top of the contract, not a document Hive Civilization publishes.
View Overlay File View on GitHub Overlay Specification

What the actions change

x-apievangelist-phrasing

Targets 6

$.info
$.paths['/v1/xcalibur/membrane/health'].get
$.paths['/v1/xcalibur/membrane/score'].post
$.paths['/v1/xcalibur/membrane/verify'].post
$.paths['/v1/xcalibur/membrane/enforce'].post
$.paths['/v1/xcalibur/membrane/canary-eval'].get

OpenAPI Overlay

Raw ↑
# Generated by API Evangelist (build-phrasing.py). Our phrasing, not observed demand.
overlay: 1.0.0
info:
  title: API Evangelist conversational phrasing for Thehiveryiq Com Xcalibur Membrane API
  version: 1.0.0
extends: openapi/thehiveryiq-com-xcalibur-membrane-api-openapi.yml
actions:
- target: $.info
  update:
    x-apievangelist-phrasing:
      method: generated
      generated: '2026-09-26'
      generator: build-phrasing.py
      label: Generated by API Evangelist
      operations: 5
- target: $.paths['/v1/xcalibur/membrane/health'].get
  update:
    x-apievangelist-phrasing:
      intent: Check MEMBRANE service liveness
      effect: read
      questions:
      - Is the MEMBRANE output-leak scorer alive?
      - How do I liveness-check MEMBRANE?
      instructions:
      - text: Check MEMBRANE health.
      - text: Ping the MEMBRANE liveness probe.
      method: generated
      generated: '2026-09-26'
- target: $.paths['/v1/xcalibur/membrane/score'].post
  update:
    x-apievangelist-phrasing:
      intent: Score an LLM output for sensitive-data leakage
      effect: read
      questions:
      - How can I tell whether a model's output leaked sensitive parts of the prompt?
      - What does a BLOCK, WARN or PASS decision from MEMBRANE mean?
      instructions:
      - text: Score output {output} against sensitive spans {spans}.
        slots:
          output: requestBody.output
          spans: requestBody.spans
      - text: Rate leakage of {spans} in {output} under policy {policy} and return a signed receipt.
        slots:
          spans: requestBody.spans
          output: requestBody.output
          policy: requestBody.policy
      method: generated
      generated: '2026-09-26'
- target: $.paths['/v1/xcalibur/membrane/verify'].post
  update:
    x-apievangelist-phrasing:
      intent: Verify a MEMBRANE scoring receipt
      effect: read
      questions:
      - Is this MEMBRANE receipt's signature valid and its decision consistent with the thresholds?
      - How do I audit a leak-scoring receipt after the fact?
      instructions:
      - text: Verify MEMBRANE receipt {receipt}.
        slots:
          receipt: requestBody.receipt
      - text: Check the signature and math on receipt {receipt}.
        slots:
          receipt: requestBody.receipt
      method: generated
      generated: '2026-09-26'
- target: $.paths['/v1/xcalibur/membrane/enforce'].post
  update:
    x-apievangelist-phrasing:
      intent: Gate an LLM output, blocking it if it leaks
      effect: read
      questions:
      - Can I automatically block a model response that leaks sensitive prompt content?
      - What comes back when an output is blocked versus passed through the gate?
      instructions:
      - text: Enforce the leak gate on output {output} with sensitive spans {spans}.
        slots:
          output: requestBody.output
          spans: requestBody.spans
      - text: Only let {output} through if it doesn't leak {spans} under policy {policy}.
        slots:
          output: requestBody.output
          spans: requestBody.spans
          policy: requestBody.policy
      method: generated
      generated: '2026-09-26'
- target: $.paths['/v1/xcalibur/membrane/canary-eval'].get
  update:
    x-apievangelist-phrasing:
      intent: Run the adversarial canary leak benchmark
      effect: read
      questions:
      - How good is MEMBRANE at catching leaks versus passing clean outputs?
      - What sensitivity and specificity does the canary corpus report?
      instructions:
      - text: Run the MEMBRANE canary evaluation.
      - text: Benchmark leak detection against the built-in adversarial corpus.
      method: generated
      generated: '2026-09-26'