> ## Documentation Index
> Fetch the complete documentation index at: https://docs.ishlabs.io/llms.txt
> Use this file to discover all available pages before exploring further.

# Delete Task

> Retire a task.

It stops appearing in the list but stays readable by id, sessions bound to it keep resolving, and its name is freed for reuse. Idempotent. Refused with a 409 while a published listing references it.



## OpenAPI

````yaml /api/openapi.public.json delete /api/v1/tasks/{task_id}
openapi: 3.1.0
info:
  contact:
    name: ish developer platform
    url: https://docs.ishlabs.io/
  description: >
    The ish Developer API: the entry point for getting human feedback on
    whatever you're building, in whichever format you need.


    Its first surface is **sessions**: you open a session for one of your people
    inside *your own* environment, submit what that person perceives
    (human-sensory observations: a rendered screen, an error message) along with
    the set of actions your environment currently offers, and they decide one
    action per turn and tell you how it felt. Your code executes the chosen
    action and loops. **The loop is decide-only: ish never locates, drives, or
    executes anything inside your environment.** You stay in control of what
    runs; we tell you what a person would do next and why.


    ## Vocabulary


    The public vocabulary is **workspace**, **person**, **environment**, and
    **session**, and the wire uses these names directly:


    - `workspace_id`: the workspace that owns the person and is billed for the  
    session (the create path is `/workspaces/{workspace_id}/sessions`).

    - `person_id`: the person the session drives. Two kinds are available to
    you:   the ones your workspace owns, and the ones published in the shared
    ish   library. Both are listed and searchable under
    `/v1/workspaces/{workspace_id}/people`, and each carries an `owner` saying
    which it is.

    - `environment_id`: a registered environment in that workspace: a named
    target   with its own operating notes and default action vocabulary.
    Register one and   reference it, or describe an ad-hoc one inline on the
    session; supply exactly   one of the two.


    **"Session" here means one API session: a single person working toward one
    task in your environment, turn by turn.** It is not the ish product's notion
    of a study session (someone sitting down with a study), and it is unrelated
    to a browser session. Sessions live under `/v1/sessions`.


    ## Authentication


    All requests use a bearer token: `Authorization: Bearer <token>`. The token
    is a **workspace API key** (prefix `ish_sk_live_`), a workspace-scoped
    machine principal minted in Settings > Developers and shown exactly once at
    mint. Scopes: `sessions:run` (create, submit turns, close), `sessions:read`
    (read the session and trace), `environments:read` / `environments:write` for
    the environment registry, `tasks:read` / `tasks:write` for the task
    registry, `people:read` / `people:write` for the people you can run a
    session for, and `usage:read` for this workspace's consumption, limits and
    rate-limit budgets. Keys are minted with `sessions:run`, `sessions:read`,
    and `tasks:read`; ask for any `:write` scope, and for `people:read` and
    `usage:read`, explicitly. `usage:read` is never implied by another scope: it
    covers the shape of the whole workspace's operation, not just the lane a key
    was minted for. An ish user access token (JWT) also works on the same header
    for personal scripts; server integrations should use keys.


    ## Versioning & forward-compatibility


    The API is versioned in the URL path (`/api/v1`). Several enum-typed and
    block-typed fields are **open unions**: new values may be added over time,
    and clients MUST tolerate values they do not recognize rather than failing
    on them. The field descriptions call these out.


    ## Correlation & errors


    Every response carries an **`X-Request-Id`** header; quote it in support
    requests. Error responses additionally echo it in the body as `request_id`.


    Two error-body shapes exist; parse defensively, they differ:


    - **`ValidationError`**: the request-validation layer (HTTP 422): a
    malformed   body (wrong types, missing fields, an invalid observation
    `type`, a bad enum   value, `parameters` that don't fit the model). A
    field-level envelope   (`error_code`, `errors[]`, `suggestions[]`;
    `suggestions` is CLI-oriented   boilerplate; ignore it for API use).

    - **`Error`**: everything else, a `{ detail, request_id }` envelope. This  
    covers the handler layer's business-rule 422s (e.g. `exactly one image  
    observation block is required`; `person has no renderable background`) as
    well   as 401 / 402 / 403 / 404 / 409 / 429 / 502.


    That is why 422 is documented on the responses below as either shape
    (`oneOf`).


    **Branch on `detail.error_kind`, not on the status code.** On every refusal
    this API classifies, `detail` is an object rather than a string:


    ```json

    {
      "detail": {
        "error_kind": "turn_index_mismatch",
        "message": "expected turn_index 4, got 3",
        "expected_turn_index": 4
      },
      "request_id": "9f2c...",
      "docs_url": "https://docs.ishlabs.io/api/errors#turn_index_mismatch"
    }

    ```


    `error_kind` is the stable, machine-readable classification; `message` is
    for humans and may be reworded; any further keys carry the specifics you
    would act on. One status often covers several kinds with different remedies.
    A 402 is either "top up" or "raise your cap"; a 429 is either "slow down" or
    "you are out of sessions for today". The status alone is not enough to
    decide what to do. `docs_url` links to that kind's entry in the error
    catalog.


    The set of kinds is an **open union**: new ones are added as new refusals
    are classified, so treat an unfamiliar `error_kind` as "an error of this
    status class" rather than failing on it. `message` is always present.


    ## Rate limits


    API keys carry a per-key request budget and, on session creates, a per-key
    daily session quota. Exceeding either is a `429` with a `Retry-After` header
    (also echoed as `detail.retry_after_seconds`) and a distinct `error_kind`:
    `rate_limited` resets within the minute, `session_quota_exceeded` on a day
    boundary. Both are transient and neither affects sessions that are already
    open.


    You do not have to wait for the `429` to find out. Every key-authed
    response, including the `429` itself, carries the budget it just spent:


    - **`x-ratelimit-limit`**: the ceiling for that window.

    - **`x-ratelimit-remaining`**: what is left in it after this request.

    - **`x-ratelimit-reset`**: **seconds from now** until the window resets. A  
    delta, not a Unix timestamp.


    When a request charges more than one budget (a session create spends both),
    the headers describe the one closest to refusing you, so pacing off
    `remaining` is always pacing off the binding constraint. Treat all three as
    optional: they are absent when nothing was charged, and `reset` alone can be
    omitted if the limit backend cannot be read at that moment.


    User access tokens are deliberately not rate limited, so their responses
    carry no `x-ratelimit-*` headers. The per-key budgets exist because a key is
    a machine principal that can loop unattended; an interactive session is
    bounded by the person driving it.
  license:
    identifier: LicenseRef-Proprietary
    name: Proprietary
  title: ish Developer API
  version: 1.6.1
servers:
  - description: Production
    url: https://api.ishlabs.io
security:
  - bearerAuth: []
tags:
  - description: >-
      Open a session for one of your people inside your own environment, submit
      what they perceive turn by turn, and read back the decision they made and
      how it felt.
    name: sessions
  - description: >-
      Register the targets your sessions run against, and the declared versions
      of each one. A registered environment carries its own operating notes and
      default action vocabulary, so a session only has to name it.
    name: environments
  - description: >-
      Register the intent a session works toward, so runs stay comparable across
      environments and over time. A session can also describe its instructions
      inline and skip the registry entirely.
    name: tasks
  - description: >-
      Find the person a session should run as. Read and search the pool you can
      draw on: the people your workspace owns, plus the ones published in the
      shared ish library. Read-only at this version.
    name: people
  - description: >-
      What this workspace has consumed, and what bounds it: the session usage
      series over time, the spend cap and the credits drawn against it, and the
      live rate-limit budgets this credential is spending. Read-only.
    name: usage
paths:
  /api/v1/tasks/{task_id}:
    delete:
      tags:
        - tasks
      summary: Delete Task
      description: >-
        Retire a task.


        It stops appearing in the list but stays readable by id, sessions bound
        to it keep resolving, and its name is freed for reuse. Idempotent.
        Refused with a 409 while a published listing references it.
      operationId: deleteTask
      parameters:
        - in: path
          name: task_id
          required: true
          schema:
            format: uuid
            title: Task Id
            type: string
      responses:
        '204':
          description: Successful Response
          headers:
            X-Request-Id:
              $ref: '#/components/headers/XRequestId'
            x-ratelimit-limit:
              $ref: '#/components/headers/XRateLimitLimit'
            x-ratelimit-remaining:
              $ref: '#/components/headers/XRateLimitRemaining'
            x-ratelimit-reset:
              $ref: '#/components/headers/XRateLimitReset'
        '401':
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/Error'
          description: >-
            The bearer token is missing, malformed, unknown, or revoked
            (`detail.error_kind = 'invalid_api_key'`). One kind covers all four
            deliberately: distinguishing them would confirm which guesses are
            real keys.
          headers:
            X-Request-Id:
              $ref: '#/components/headers/XRequestId'
            x-ratelimit-limit:
              $ref: '#/components/headers/XRateLimitLimit'
            x-ratelimit-remaining:
              $ref: '#/components/headers/XRateLimitRemaining'
            x-ratelimit-reset:
              $ref: '#/components/headers/XRateLimitReset'
        '403':
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/Error'
          description: >-
            The API key lacks the `tasks:write` scope (`detail.error_kind =
            'insufficient_scope'`, with `detail.required_scope =
            'tasks:write'`). Mint a key that includes it; an ish user access
            token is unconstrained by scopes.
          headers:
            X-Request-Id:
              $ref: '#/components/headers/XRequestId'
            x-ratelimit-limit:
              $ref: '#/components/headers/XRateLimitLimit'
            x-ratelimit-remaining:
              $ref: '#/components/headers/XRateLimitRemaining'
            x-ratelimit-reset:
              $ref: '#/components/headers/XRateLimitReset'
        '404':
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/Error'
          description: >-
            The task was not found, or is not reachable with this credential.
            `detail.error_kind` is `task_not_found`.
          headers:
            X-Request-Id:
              $ref: '#/components/headers/XRequestId'
            x-ratelimit-limit:
              $ref: '#/components/headers/XRateLimitLimit'
            x-ratelimit-remaining:
              $ref: '#/components/headers/XRateLimitRemaining'
            x-ratelimit-reset:
              $ref: '#/components/headers/XRateLimitReset'
        '409':
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/Error'
          description: >-
            An active recruitment listing references this task
            (`detail.error_kind = 'task_in_published_listing'`, with
            `detail.listing_id`). Close the listing, or remove the task from it,
            then retry.
          headers:
            X-Request-Id:
              $ref: '#/components/headers/XRequestId'
            x-ratelimit-limit:
              $ref: '#/components/headers/XRateLimitLimit'
            x-ratelimit-remaining:
              $ref: '#/components/headers/XRateLimitRemaining'
            x-ratelimit-reset:
              $ref: '#/components/headers/XRateLimitReset'
        '422':
          content:
            application/json:
              schema:
                oneOf:
                  - $ref: '#/components/schemas/ValidationError'
                  - $ref: '#/components/schemas/Error'
          description: >-
            Validation error. A request-shape failure uses the `ValidationError`
            envelope; a handler-raised domain-validation refusal uses the
            `Error` envelope.
          headers:
            X-Request-Id:
              $ref: '#/components/headers/XRequestId'
            x-ratelimit-limit:
              $ref: '#/components/headers/XRateLimitLimit'
            x-ratelimit-remaining:
              $ref: '#/components/headers/XRateLimitRemaining'
            x-ratelimit-reset:
              $ref: '#/components/headers/XRateLimitReset'
        '429':
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/Error'
          description: >-
            The API key's request rate limit is exceeded (`detail.error_kind =
            'rate_limited'`). Transient: retry after the `Retry-After` interval,
            also echoed as `detail.retry_after_seconds`. Not applied to user
            access tokens.
          headers:
            X-Request-Id:
              $ref: '#/components/headers/XRequestId'
            x-ratelimit-limit:
              $ref: '#/components/headers/XRateLimitLimit'
            x-ratelimit-remaining:
              $ref: '#/components/headers/XRateLimitRemaining'
            x-ratelimit-reset:
              $ref: '#/components/headers/XRateLimitReset'
      security:
        - bearerAuth: []
components:
  headers:
    XRequestId:
      description: >-
        Per-response correlation id, present on EVERY response (success and
        error). Error bodies echo it as `request_id`.
      schema:
        type: string
    XRateLimitLimit:
      description: >-
        Ceiling for the rate-limit window this request charged. Present on
        responses made with an API key against a rate-limited operation,
        including the `429` itself; absent for user access tokens (which carry
        no per-key budget) and when the rate-limit backend could not be read.
      schema:
        type: integer
    XRateLimitRemaining:
      description: >-
        Requests left in that window AFTER this one was counted. Pace off this
        rather than waiting for a `429`.
      schema:
        type: integer
    XRateLimitReset:
      description: >-
        **Seconds from now** until that window resets (a delta, NOT a Unix
        timestamp). Measured when the response is written. Omitted on the rare
        occasion the reset instant could not be read, while `limit` and
        `remaining` are still published.
      schema:
        type: integer
  schemas:
    Error:
      description: >-
        Generic error envelope for 4xx/5xx responses other than request-shape
        validation.
      properties:
        detail:
          anyOf:
            - type: string
            - additionalProperties: true
              type: object
          description: >-
            A human-readable message, OR a structured object carrying a
            machine-readable `error_kind` (e.g. `idempotency_conflict`,
            `turn_index_mismatch`, `decision_invalid`).
          title: Detail
        docs_url:
          description: >-
            Link to this error's entry in the error catalog. Present whenever
            `detail` carries a registered `error_kind`; absent on unclassified
            errors.
          title: Docs Url
          type: string
        request_id:
          description: >-
            Correlation id; identical to the `X-Request-Id` response header.
            Quote it in support requests.
          title: Request Id
          type: string
      required:
        - detail
        - request_id
      title: Error
      type: object
    ValidationError:
      description: >-
        Request body / parameter validation failure (HTTP 422). A field-level
        envelope so callers can pinpoint the offending field, value, and allowed
        values without parsing prose.
      properties:
        detail:
          title: Detail
          type: string
        error:
          title: Error
          type: string
        error_code:
          description: Stable machine code, e.g. `validation_error`.
          title: Error Code
          type: string
        errors:
          items:
            $ref: '#/components/schemas/FieldError'
          title: Errors
          type: array
        retryable:
          title: Retryable
          type: boolean
        status:
          title: Status
          type: integer
        suggestions:
          items:
            type: string
          title: Suggestions
          type: array
      required:
        - error
        - error_code
        - status
        - retryable
        - detail
        - errors
      title: ValidationError
      type: object
    FieldError:
      description: One field-level entry in a ValidationError envelope.
      properties:
        allowed_values:
          description: For enum/literal failures, the accepted values.
          items: {}
          title: Allowed Values
          type: array
        input:
          anyOf:
            - type: string
            - type: number
            - type: boolean
            - type: 'null'
          description: Offending value (truncated); omitted for containers.
          title: Input
        loc:
          description: Path to the offending field.
          items:
            anyOf:
              - type: string
              - type: integer
          title: Location
          type: array
        msg:
          title: Message
          type: string
        type:
          title: Error Type
          type: string
      required:
        - loc
        - msg
        - type
      title: FieldError
      type: object
  securitySchemes:
    bearerAuth:
      description: >-
        Workspace API key as a bearer token: `Authorization: Bearer
        ish_sk_live_...`. Keys are workspace-scoped machine principals minted in
        Settings > Developers (shown once at mint). Scopes: `sessions:run`
        (create a session, submit turns, close), `sessions:read` (read the
        session and its decision trace), `environments:read` and
        `environments:write` (manage the workspace's registered environments),
        `tasks:read` and `tasks:write` (manage the workspace's registered
        tasks), `people:read` and `people:write` (the people the workspace can
        run a session for), and `usage:read` (the workspace's own consumption,
        spend limits and rate-limit budgets). A key is minted with
        `sessions:run`, `sessions:read`, and `tasks:read` by default; every
        other scope must be requested, `people:read` and `usage:read` included.
        An ish user access token also works on the same header for personal
        scripts.
      scheme: bearer
      type: http

````