generated: '2026-09-19'
method: derived
source: >-
  Derived by aligning the nine skills in the A2A agent card (a2a/phoslabs-io-agent-card.json), the six tools
  in the provider's live REST listing (GET https://phoslabs.io/api/v1/tools, saved as
  plans/phoslabs-io-api-v1-tools.json) and the seven operations in the provider-published OpenAPI
  (openapi/phoslabs-io-openapi.yml). The hosted MCP server's tools/list is OAuth-gated and was not
  introspected, so MCP tool names are NOT known: rows are keyed on the agent-card skill id, which is the
  provider's own identifier for the same capability, and confidence is set accordingly.
purpose: >-
  Record how Phos Labs' three agent-facing projections — the OAuth-gated MCP server, the A2A-shaped discovery
  card and the Bearer-key REST API — relate. They describe one product but are not the same surface: the REST
  contract is the narrowest (7 operations), the card is the widest (9 skills), and the MCP inventory is
  unverifiable anonymously.
surfaces:
  rest_openapi: openapi/phoslabs-io-openapi.yml
  rest_openapi_note: 7 operations on https://phoslabs.io/api/v1 (6 POST tools + anonymous GET /api/v1/tools). All six POST tools answer 401 without a key.
  rest_live_listing: https://phoslabs.io/api/v1/tools
  mcp: https://mcp.phoslabs.io/mcp
  mcp_note: 'tools/list GATED (401 invalid_token, OAuth 2.1 per RFC 9728/8414 metadata). Root path answers initialize only; /sse wants an API key.'
  a2a: null
  a2a_note: The agent card's url is the MCP host; message/send returns 401 "use /mcp for full MCP access". No A2A transport.
  graphql: null
crosswalk:
- tool: diagnose-dropoff
  category: conversion
  rest: [diagnose]
  binding: rest
  confidence: medium
  note: >-
    Card skill "Diagnose Customer Drop-off" and REST diagnose (POST /api/v1/diagnose, 10 credits) describe the
    same job on the same input (a funnel description plus optional analytics data). The inherited input
    contract is diagnose's requestBody: funnel_description (required), data (object), industry.
- tool: fix-checkout
  category: conversion
  rest: [fixCheckout]
  binding: rest
  confidence: medium
  note: 'Same id on both sides; POST /api/v1/fix-checkout, 30 credits; requestBody checkout_flow (required).'
- tool: write-product-copy
  category: copywriting
  rest: [copy]
  binding: rest
  confidence: medium
  note: 'Card skill lists social proof, loss framing, anchoring, scarcity; REST copy says the same principles. POST /api/v1/copy, 20 credits; requestBody draft + goal (both required).'
- tool: optimize-pricing
  category: pricing
  rest: [pricing]
  binding: rest
  confidence: medium
  note: 'Both name anchoring and decoy pricing. POST /api/v1/pricing, 25 credits; requestBody product (required), context, persona.'
mcp_only:
- tool: predict-churn
  reason: No REST operation in the OpenAPI or the live /api/v1/tools listing; the README names /churn (15 credits) but it is not served in the contract.
- tool: personalize-approach
  reason: No REST operation; segmentation by decision style appears only in the card.
- tool: add-social-proof
  reason: README names /social-proof (10 credits); absent from the OpenAPI and the live listing.
- tool: run-experiment
  reason: README names /experiment (15 credits); absent from the OpenAPI and the live listing.
- tool: ethics-check
  reason: >-
    No REST operation. detectBiases (POST /api/v1/detect-biases) is the nearest REST neighbour but audits the
    caller's own text for cognitive biases, not a design for dark patterns; not bound.
rest_only:
- capability: free full audit
  operations: [audit]
  note: POST /api/v1/audit, 0 credits per the live listing (the README says 80). No card skill.
- capability: cognitive-bias detection
  operations: [detectBiases]
  note: POST /api/v1/detect-biases, 8 credits. No card skill (see ethics-check above).
- capability: tool discovery
  operations: [listTools]
  note: GET /api/v1/tools, anonymous (security []). Returns the REST tool inventory with prices and input field lists.
coverage:
  card_skills: 9
  card_skills_bound_to_rest: 4
  card_skills_mcp_only: 5
  rest_operations: 7
  rest_operations_with_a_skill: 4
  rest_only: 3
  mcp_tools_named_by_server: 0
  mcp_tools_introspected: false