generated: '2026-08-13'
method: generated
source: >-
  openapi/usergems-*-openapi.yml grounded against
  https://app.usergems.com/api/documentation and
  https://help.usergems.com/article/using-the-usergems-api
description: >-
  Packaged Agent Skills for the UserGems REST API. Every operationId referenced
  below exists verbatim in the linked spec. UserGems publishes no AGENTS.md or
  first-party skill files, so these are API Evangelist generations; they encode
  the conventions, destructive-default warnings and retry rules captured in
  conventions/ and errors/ so an agent calling this write-only API does not learn
  them the hard way.
skills:
  - file: usergems-track-champion-contacts.md
    name: Track champion contacts in UserGems
    api: openapi/usergems-contacts-api-openapi.yml
    operations: [addContact, deleteContact]
    consequence: write
  - file: usergems-manage-target-accounts.md
    name: Manage UserGems target accounts
    api: openapi/usergems-accounts-api-openapi.yml
    operations: [addAccount, deleteAccount]
    consequence: write
  - file: usergems-honor-erasure-request.md
    name: Honor a UserGems data-subject erasure request
    api: openapi/usergems-privacy-api-openapi.yml
    operations: [privacyDelete]
    consequence: irreversible
    human_in_the_loop: recommended
coverage:
  operations_total: 5
  operations_covered: 5
notes:
  - >-
    No skill is written for the UserGems MCP server: its tool names and input
    schemas are auth-gated (tools/list returns 401), and grounding a skill in
    capability labels rather than real tool identifiers would teach an agent to
    call tools that may not exist under those names.