generated: '2026-09-04'
method: probed
source: https://docs.shieldlabs.ai/.well-known/agent-card.json
description: >-
  A2A Agent Card observed at /.well-known/agent-card.json on docs.shieldlabs.ai. Provider-published
  by construction — served from a ShieldLabs-controlled host; nothing here is derived or generated.
  Graded against the A2A 1.0.0 AgentCard object.

  RE-PROBED 2026-09-04: still HTTP 200, still application/json, and the card CHANGED. The provider
  rewrote the description of its one advertised skill — it is now organised around the situations an
  agent finds itself in (building login/signup flows, processing payments, protecting sensitive
  actions, analysing traffic quality) rather than around product features. The card's structure,
  version, protocolVersion and skill URL are unchanged, so the grade does not move. The new body is
  saved verbatim over the old one.
discovery:
  path: /.well-known/agent-card.json
  canonical: true
  host: docs.shieldlabs.ai
  notes: >-
    Probed on every ShieldLabs host. shieldlabs.ai, www.shieldlabs.ai, api.shieldlabs.ai and
    account.shieldlabs.ai all return 404 for both /.well-known/agent-card.json and the legacy
    /.well-known/agent.json; re-probed 2026-09-04 across eight hosts with the same result. The one
    host that answers 200 for those paths, app.shieldlabs.ai, answers 200 for a negative-control path
    that cannot exist, so it is a single-page-app catch-all and not a card. Only the documentation host serves a card, and it is emitted by the
    docs platform (Mintlify, x-matched-path /_sites/[subdomain]/.well-known/agent-card.json)
    rather than by the ShieldLabs Server API itself — the card's url, provider.url and
    documentationUrl all point back at https://docs.shieldlabs.ai/, not at the API surface.
conformance:
  spec: A2A 1.0.0
  grade: flavored
  protocol_version: '0.3'
  preferred_transport: HTTP+JSON
  hard_checks:
    capabilities_is_object: true
    protocol_version_present: true
    skills_is_array: true
  deviations:
  - supportedInterfaces-instead-of-additionalInterfaces
  - protocol-version-0.3-pre-1.0
  grade_rationale: >-
    All three hard checks pass — capabilities is an object, protocolVersion is present, skills is an
    array — but the card uses the pre-1.0 `supportedInterfaces` key in place of
    `additionalInterfaces` and declares protocolVersion 0.3, so it is graded flavored rather than
    conformant. Unchanged on the 2026-09-04 re-probe.
card:
  name: ShieldLabs
  description: >-
    Visitor identification and abuse-prevention docs for ShieldLabs: snippet, Risk Score, anonymity
    signals, Patterns, webhooks, and Server API.
  url: https://docs.shieldlabs.ai/
  version: 1.0.0
  organization: ShieldLabs
  documentation_url: https://docs.shieldlabs.ai/
  capabilities:
    streaming: false
    push_notifications: false
  default_input_modes:
  - text/plain
  default_output_modes:
  - text/plain
  skills: 1
  skill_ids:
  - shieldlabs
  skill_urls:
  - https://docs.shieldlabs.ai/.well-known/agent-skills/shieldlabs/skill.md
  file: shieldlabs-agent-card.json
x-evidence:
  fetched: '2026-09-04'
  url: https://docs.shieldlabs.ai/.well-known/agent-card.json
  http_status: 200
  content_type: application/json
  bytes: 1202
  negative_control: /.well-known/shieldlabs-negative-control-7f3ab91c.json returned 404 on this host
  first_fetched: '2026-08-19'
  changed_since_first_fetch: skills[0].description rewritten by the provider
  note: Verbatim body saved alongside this manifest. No field inferred.