generated: '2026-09-07'
method: generated
source: >-
  Grounded in openapi/plumma-connect-openapi.yml — the single operationId `connectApi` was
  verified against the spec; no operation, field or command name below was invented. Runtime
  rules come from conventions/, errors/, sandbox/, plans/ and rate-limits/ in this repo,
  each of which cites the provider document it was read from.
note: >-
  This API has ONE operation, so skills are organised by the flow an agent is actually
  running rather than by endpoint. Each one is the same POST /api call with a different
  `commands` payload and, more importantly, a different way of reading the answer.
skills:
- file: plumma-sim-swap-fraud-check.md
  name: SIM-swap fraud check
  api: openapi/plumma-connect-openapi.yml
  operations: [connectApi]
  commands: [sim_swap]
- file: plumma-kyc-match.md
  name: KYC match against operator records
  api: openapi/plumma-connect-openapi.yml
  operations: [connectApi]
  commands: [kyc_match]
- file: plumma-multi-command-partial-reply.md
  name: Multi-command calls and partial replies
  api: openapi/plumma-connect-openapi.yml
  operations: [connectApi]
  commands: [current_carrier, line_classification, sim_swap]
excluded:
- skill: number_verification / scam_check / device_location
  reason: >-
    Documented by the provider and present in the cmd_enc ordinal table, but absent from the
    OpenAPI allowedCommandValues enum, so a client built from the contract cannot send them.
    Writing a skill around a command the contract rejects would teach an agent to make a call
    that fails. Recorded in vocabulary/plumma-commands-vocabulary.yml instead.