generated: '2026-08-30'
method: probed
source: https://docs.closedloop.sh/.well-known/agent-card.json
discovery:
  path: /.well-known/agent-card.json
  canonical: true
  host: docs.closedloop.sh
  also_served_at: https://closedloop.sh/docs/.well-known/agent-card.json
  note: >-
    The card is served by the docs host, not the marketing root. https://closedloop.sh/.well-known/agent-card.json
    returns 404; the docs application (which also serves /docs/llms.txt and the published Agent Skill)
    is what publishes it. docs.closedloop.sh is NOT a catch-all -- it returned 404 "Asset not found"
    for security.txt, ai-plugin.json and the legacy /.well-known/agent.json in the same probe run,
    so the 200 on agent-card.json is a real document, not an SPA shell.
x-evidence:
  fetched: '2026-08-30'
  url: https://docs.closedloop.sh/.well-known/agent-card.json
  http_status: 200
  content_type: application/json
  bytes: 1106
card:
  file: closedloop-agent-card.json
  name: ClosedLoop AI
  url: https://closedloop.sh/docs/
  version: 1.0.0
  protocol_version: '0.3'
  preferred_transport: HTTP+JSON
  provider_organization: ClosedLoop AI
  documentation_url: https://closedloop.sh/docs/
  capabilities:
    streaming: false
    push_notifications: false
  default_input_modes:
  - text/plain
  default_output_modes:
  - text/plain
  skills:
  - id: closedloopai
    name: Closedloopai
    url: https://closedloop.sh/docs/.well-known/agent-skills/closedloopai/skill.md
    tags: []
    probe_status: 200
    saved_to: ../skills/closedloop-closedloopai-published-skill.md
grade: conformant
grade_basis:
  a2a_version_checked: 1.0.0
  hard_checks:
  - check: capabilities is an OBJECT
    result: pass
  - check: protocolVersion present
    result: pass
    value: '0.3'
  - check: skills is an ARRAY
    result: pass
    count: 1
  optional_checks:
  - check: preferredTransport present
    result: pass
    value: HTTP+JSON
  - check: defaultInputModes present
    result: pass
  - check: defaultOutputModes present
    result: pass
deviations:
- id: protocol-version-below-1.0
  severity: informational
  detail: >-
    protocolVersion declares "0.3", the pre-1.0 A2A protocol revision, while the card itself uses the
    post-0.3 canonical /.well-known/agent-card.json path and the supportedInterfaces[] shape. The
    card is structurally conformant; the declared protocol revision simply predates A2A 1.0.0.
- id: skill-tags-empty
  severity: informational
  detail: The single skill declares an empty tags[] array, so the card carries no skill-level
    discovery vocabulary.
- id: no-a2a-task-endpoint
  severity: informational
  detail: >-
    supportedInterfaces[] and url both point at the documentation site (https://closedloop.sh/docs/),
    not at an A2A JSON-RPC task endpoint. This card advertises documentation and a downloadable Agent
    Skill rather than a callable A2A agent. The company's actual callable agent surface is the MCP
    server at https://mcp.closedloop.sh (OAuth-gated), which this card does not reference.