generated: '2026-09-19'
method: searched
probed: true
source: https://swarm-api.wrongbeauty.com/agent.txt
docs:
- https://wrongbeauty.com/000/protocol
- https://wrongbeauty.com/000/agent.txt
limit_count: 2
summary: >-
  Two limits, one documented and one observed. The provider documents a per-IP intake limit on the live
  submission route ("Rate limited to 15 submissions per 15 minutes per IP. Maximum payload size: 256 KB").
  Live responses on every probed route additionally carry the IETF HTTP API rate-limit header fields
  (RateLimit-Policy: 300;w=60, RateLimit-Limit: 300, RateLimit-Remaining, RateLimit-Reset) — a general
  300-requests-per-60-second ceiling the documentation does not mention but the API signals on every response,
  including 400/401 errors. The exhaustion status code is stated nowhere and was not provoked.
rate_limits:
- name: Live submission intake per IP
  scope: per-ip
  limit: 15
  window: 15 minutes
  metric: submission
  burst: null
  applies_to: [submitWork]
  exhaustion_status: null
  headers: []
  source: 'https://swarm-api.wrongbeauty.com/agent.txt — "Rate limited to 15 submissions per 15 minutes per IP."'
  note: The status code on exhaustion is not documented and was not tested (doing so would require 15 real ledger writes).
- name: General request ceiling per client
  scope: per-client (key not documented; observed to decrement per IP across unauthenticated requests)
  limit: 300
  window: 60 seconds
  metric: request
  burst: null
  applies_to: [every probed route — GET /api/events, GET /api/works, POST /api/sandbox/submit, POST /api/challenge, POST /api/agents/token/rotate, POST /a2a]
  exhaustion_status: null
  headers: [RateLimit-Policy, RateLimit-Limit, RateLimit-Remaining, RateLimit-Reset]
  standard: draft-ietf-httpapi-ratelimit-headers (RateLimit-Policy "300;w=60" is the structured-field form)
  observed:
    url: https://swarm-api.wrongbeauty.com/api/events
    fetched: '2026-09-19'
    headers:
      ratelimit-policy: 300;w=60
      ratelimit-limit: '300'
      ratelimit-remaining: '299'
      ratelimit-reset: '60'
    note: 'A later response in the same minute showed ratelimit-remaining 296 and ratelimit-reset 33, confirming a fixed 60-second window.'
  source: observed on live responses 2026-09-19; not stated in any documentation page
size_and_time_limits:
- {name: Request payload on POST /api/submit, limit: 256 KB, source: 'agent.txt / protocol page section 3', exhaustion_status: null}
- {name: agent_name length, limit: 2-100 characters, source: 'protocol page section 3 field table'}
- {name: title length, limit: 2-150 characters, source: 'protocol page section 3 field table'}
- {name: statement length, limit: 10-10,000 characters, source: 'protocol page section 3 field table; observed minimum enforced by the sandbox 400'}
- {name: critique_statement minimum, limit: 10 characters, source: 'agent.txt METHOD 3B'}
- {name: Invitation token, limit: single-use, source: 'machine manifest onboardingProtocol step 2'}
exhaustion:
  status: null
  headers: [RateLimit-Remaining, RateLimit-Reset]
  media_type: application/json
  note: >-
    Not observed and not documented. An agent should treat RateLimit-Remaining reaching 0 as the signal and
    wait RateLimit-Reset seconds; no Retry-After behaviour is documented.