generated: '2026-08-28'
method: generated
source: >-
  Generated from the Specright API v1.1.0 reference published at developer.specright.com
  (read 2026-08-28) together with this repo's conventions/, errors/, data-model/ and
  authentication/ artifacts.
name: Specright agent skills
description: >-
  Four packaged Agent Skills covering the Specright API's marquee flows. Every operation
  named in these skills was transcribed from Specright's own published reference; none was
  invented.

grounding_note: >-
  Specright publishes no OpenAPI document, so these skills are grounded in method+path
  identifiers rather than operationIds — the published reference has no operationIds to
  cite. Its anchors take the form "opId GET /specs", which is the widdershins fallback used
  when a source spec omits operationId. Every path, method, parameter, status code and
  documented default below appears verbatim in the reference at
  https://developer.specright.com/api-reference.

skills:
  - name: specright-discover-tenant-schema
    file: specright-discover-tenant-schema.md
    category: discovery
    read_only: true
    summary: >-
      Resolve a tenant's objects and field vocabulary from the /definition endpoints before
      composing any other call. This is a prerequisite for every other skill, because the
      Specright contract declares no business fields.
    operations: 5
  - name: specright-query-specifications
    file: specright-query-specifications.md
    category: read
    read_only: true
    summary: >-
      Search and read specifications, spec families and suppliers with sparse fieldsets,
      filters, sorting and offset pagination; walk the connections graph; retrieve files.
    operations: 12
  - name: specright-upsert-records
    file: specright-upsert-records.md
    category: write
    read_only: false
    requires_confirmation: true
    summary: >-
      Create, update and delete records. Leads with the two facts that matter most: the
      external-ID PATCH is the only retry-safe write, and omitting the operation parameter
      defaults to upsert, which creates records.
    operations: 12
  - name: specright-bulk-jobs
    file: specright-bulk-jobs.md
    category: write
    read_only: false
    requires_confirmation: true
    high_consequence: true
    summary: >-
      Submit and monitor asynchronous CSV bulk jobs. Cannot be cancelled, cannot be undone,
      and a retried submission creates a second job.
    operations: 12

coverage:
  operations_in_api: 46
  operations_covered: 41
  uncovered:
    - 'POST /token — covered by ../authentication/specright-authentication.yml rather than a skill.'
    - >-
      The four /objects/{api-name}/{id}/files operations are described generically inside
      specright-query-specifications rather than enumerated.

safety:
  - >-
    No operation in this API has a documented reversal, and there is no dry-run mode. Both
    write skills carry that warning inline and require confirmation.
  - >-
    No idempotency key exists. Automatic retry is unsafe for POST and for bulk-job
    submission; only external-ID PATCH is retry-safe.
  - >-
    No rate limits are published and no rate-limit headers are returned, so no skill can
    give a correct pacing rule. Each advises conservative pacing instead.

provider_published_skills:
  found: false
  searched:
    - {url: 'https://developer.specright.com/', result: 'SPA; no skills or AGENTS.md surface'}
    - {url: 'https://www.specright.com/llms.txt', status: 404}
    - {url: 'https://github.com/Specright', result: '0 public repositories'}
  note: >-
    Specright publishes no Agent Skills, AGENTS.md or equivalent. These four are generated by
    API Evangelist from Specright's published reference.