generated: '2026-09-04'
method: searched
source: >-
  registry.npmjs.org search, pypi.org, api.github.com/orgs/workera-ai,
  https://skills.workera.ai/api/docs
official_sdk_count: 0
community_sdk_count: 0
packages: []
findings:
  npm:
    searched: 'https://registry.npmjs.org/-/v1/search?text=workera'
    result: no Workera-published package; every hit is an unrelated "worker*" package
  pypi:
    probed: https://pypi.org/pypi/workera/json
    status: 404
  other_registries:
    checked:
    - Maven Central
    - NuGet
    - RubyGems
    - crates.io
    - pkg.go.dev
    result: not checked individually — the docs and the GitHub org name no client library in any language, so there is no package name to resolve.
  github_org:
    url: https://github.com/workera-ai
    public_repos: 13
    first_party_client_libraries: 0
    note: >-
      All 13 public repositories are forks of third-party CI actions and Elixir libraries used to build
      the product (elixir_samly, kaffy, langfuse_sdk, instructor_ex, openai.ex, calver-tag-action,
      git-auto-commit-action, files-sync-action, merge-branch, action-semantic-pull-request,
      release-action, changelog-action, airtable.py); seven are archived. Nothing published for API
      consumers. The org confirms the platform is an Elixir/Phoenix application, which the API's
      operationIds (WorkeraWebappsWeb.Rest.Controllers.*) also show.
  docs:
    url: https://skills.workera.ai/api/docs
    sdk_section: absent
    note: >-
      The reference documents raw HTTP only. Its single code example is a curl command; there is no
      install snippet, no language tab and no SDK link anywhere in the document.
assessment: >-
  Workera ships no client SDK in any language. Integration is raw HTTPS with a bearer key. No `SDKs`
  pointer is emitted in apis.yml because there is no SDK to point at — recording an honest zero rather
  than crediting a package that does not exist. For an enterprise API whose consumers are HR and
  security engineering teams, even one maintained client would materially lower the integration cost;
  as it stands there is also no abandonment signal to read, because there is nothing published to go
  stale.