generated: '2026-08-17'
method: generated
source: >-
  Derived from openapi/ (six per-tag specs refined from
  https://api.cloud.quandela.com/openapi.json) plus
  conventions/quandela-conventions.yml, errors/quandela-problem-types.yml,
  authentication/quandela-authentication.yml,
  rate-limits/quandela-rate-limits.yml and sandbox/quandela-sandbox.yml.
note: >-
  Four packaged Agent Skills, one per marquee flow on the Quandela Cloud API.
  Every operationId referenced in these skills exists verbatim in the specs
  under openapi/ (the ids are derived — Quandela ships none — and the derivation
  is recorded in overlays/quandela-cloud-overlay.yaml). No provider-published
  skill or AGENTS.md exists: github.com/Quandela has no AGENTS.md or skills
  directory in any of its 15 repositories, and the Perceval documentation
  publishes no agent guidance. These are generated operating instructions, not
  vendor material.
skills:
  - file: quandela-submit-perceval-job.md
    name: Submit a Perceval job to a Quandela QPU and collect the result
    api: openapi/quandela-perceval-job-openapi.yml
    flow: job-execution
    operations:
      - post_api_tokens
      - get_api_jobs_availability
      - post_api_jobs
      - get_api_jobs_by_job_id_status
      - get_api_jobs_by_job_id_result
      - get_api_jobs_by_job_id_data
      - post_api_jobs_by_job_id_cancel
      - post_api_jobs_by_job_id_rerun
      - post_api_tokens_revoke
  - file: quandela-run-quantum-toolbox-algorithm.md
    name: Estimate then run a Quandela Quantum Toolbox algorithm
    api: openapi/quandela-quantum-toolbox-openapi.yml
    flow: quantum-primitives
    operations:
      - post_qt_chemistryvqe_estimate
      - post_qt_chemistryvqe
      - get_qt_chemistryvqe_by_job_id_status
      - get_qt_chemistryvqe_by_job_id_results
      - post_qt_customvqe_estimate
      - post_qt_customvqe
      - post_qt_cvarvqe_estimate
      - post_qt_cvarvqe
      - post_qt_graphDSI_estimate
      - post_qt_graphDSI
      - post_qt_graphIsomorphism_estimate
      - post_qt_graphIsomorphism
      - get_api_jobs_availability
  - file: quandela-manage-cloud-job-tokens.md
    name: Mint, rotate, audit and revoke Quandela Cloud Job Tokens
    api: openapi/quandela-perceval-job-token-openapi.yml
    flow: credential-lifecycle
    operations:
      - post_api_tokens
      - get_api_tokens
      - get_api_tokens_by_token_id
      - put_api_tokens_by_token_id
      - post_api_tokens_revoke
      - post_api_tokens_reopen
      - post_api_tokens_delete_by_ids
      - get_api_tokens_usage_report
  - file: quandela-draw-certified-quantum-randomness.md
    name: Draw certified quantum randomness from Quandela Entropy (QRNG)
    api: openapi/quandela-qrng-openapi.yml
    flow: qrng
    operations:
      - get_qt_qrng_bytes
      - get_qt_qrng_ints
cross_cutting_rules_every_skill_carries:
  - >-
    No idempotency exists on this API. `process_id` on POST /api/jobs is a
    uniqueness guard that 400s on collision, not an idempotent replay, and the
    Quantum Toolbox has no equivalent at all — so a blind retry after a timeout
    can consume credits twice.
  - >-
    HTTP 401 is overloaded. Quandela returns 401 for credit and platform-time
    exhaustion as well as for authentication failure, so an agent that handles
    401 by refreshing its credential will loop forever on "Not enough credits".
    Every skill branches on the `error` string.
  - >-
    No rate-limit headers and no 429. Concurrency ceilings surface as 400/403 and
    the concrete numbers are only available from GET /api/jobs/availability.
  - >-
    Errors are a vendor {detail, error} envelope with no machine-readable code —
    string matching on `error` is unavoidable.
  - >-
    Check GET /api/platforms/public (anonymous, undocumented) before submitting:
    on 2026-08-17 both published QPUs reported unreachable / maintenance.