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

# Evaluate an instruction proposal

> Run the bounded holdout pair: baseline and candidate replay the source Outcome's own task and rubric in two ordinary isolated Sessions, one graded iteration each, and their verdicts reconcile from those Sessions' event logs. Candidate text reaches exactly one place before apply — each holdout Session's own graded context. Requires a `drafted` or `evaluation_failed` proposal plus the requester Principal for the evaluation Sessions. Returns the InstructionProposal in `evaluating`; accepts an optional Idempotency-Key.

Checkfu support posture: alpha; hosted. Required evidence journey: skill-delivery. Deployment-specific readiness and the latest proven release are available from GET /v1/support/capabilities.



## OpenAPI

````yaml /openapi.json post /v1/instruction-proposals/{id}/evaluate
openapi: 3.1.0
info:
  title: Checkfu API
  version: '2026-08-27'
  description: >-
    Authentication is declared per operation: API-key, runtime, or connector
    bearer; Automation signature; or credential-free pairing redemption. Every
    general Checkfu REST request requires the dated `Checkfu-Version` header
    (one of: 2026-08-27); the three MCP JSON-RPC transports use
    `MCP-Protocol-Version`, A2A uses `A2A-Version`, and the provider OAuth
    callback carries neither Checkfu header. API keys resolve one Workspace
    without a request selector; authenticated responses identify it with
    `Checkfu-Workspace-Id`.
servers:
  - url: https://api.checkfu.com
security:
  - bearerAuth: []
tags:
  - name: organizations
  - name: sourceRepositories
  - name: tenants
  - name: workspaces
  - name: principals
  - name: principalGroups
  - name: principalAccessCredentials
  - name: apiKeys
  - name: agents
  - name: harnesses
  - name: harnessRuntime
  - name: sandboxProfiles
  - name: computerProfiles
  - name: permissionAssignments
  - name: actionPolicies
  - name: files
  - name: memoryStores
  - name: dreams
  - name: modelCredentials
  - name: modelRoutingProfiles
  - name: blueprintInstallations
  - name: toolSources
  - name: skills
  - name: skillSources
  - name: agentSources
  - name: skillProposals
  - name: instructionProposals
  - name: catalog
  - name: concepts
  - name: support
  - name: connections
  - name: connectionVaults
  - name: connectionAssignments
  - name: connectedRuntimes
  - name: projects
  - name: collaboration
  - name: automationGraphs
  - name: automations
  - name: actionApprovals
  - name: standingApprovals
  - name: usage
  - name: models
  - name: outcomes
  - name: budgets
  - name: billing
  - name: sessions
  - name: audit
  - name: sessionExports
  - name: runs
  - name: runnerPools
  - name: transcripts
  - name: sessionWatches
  - name: webhookEndpoints
  - name: integrationGateway
  - name: agentDeployments
  - name: workEnvironments
  - name: computers
  - name: computerScreens
  - name: computerBrowserObservations
  - name: computerBrowserActions
  - name: environments
  - name: vaults
  - name: apiMcp
  - name: a2a
paths:
  /v1/instruction-proposals/{id}/evaluate:
    post:
      tags:
        - instructionProposals
      summary: Evaluate an instruction proposal
      description: >-
        Run the bounded holdout pair: baseline and candidate replay the source
        Outcome's own task and rubric in two ordinary isolated Sessions, one
        graded iteration each, and their verdicts reconcile from those Sessions'
        event logs. Candidate text reaches exactly one place before apply — each
        holdout Session's own graded context. Requires a `drafted` or
        `evaluation_failed` proposal plus the requester Principal for the
        evaluation Sessions. Returns the InstructionProposal in `evaluating`;
        accepts an optional Idempotency-Key.


        Checkfu support posture: alpha; hosted. Required evidence journey:
        skill-delivery. Deployment-specific readiness and the latest proven
        release are available from GET /v1/support/capabilities.
      operationId: instructionProposals.evaluateInstructionProposal
      parameters:
        - name: id
          in: path
          schema:
            $ref: '#/components/schemas/InstructionProposalId'
          required: true
        - name: checkfu-version
          in: header
          schema:
            type: string
            enum:
              - '2026-08-27'
          required: true
        - name: idempotency-key
          in: header
          schema:
            type: string
            allOf:
              - maxLength: 255
          required: false
      requestBody:
        content:
          application/json:
            schema:
              type: object
              properties:
                principal:
                  $ref: '#/components/schemas/PrincipalId'
              required:
                - principal
              additionalProperties: false
        required: true
      responses:
        '200':
          description: >-
            distinct reflection and judge bindings, an error exactly on
            authoring failure, an artifact exactly past authoring, an
            application exactly on applied, a published version only after
            application, and evaluation states only with a recorded comparison
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/InstructionProposalConsistency_1'
          headers:
            checkfu-workspace-id:
              description: The Workspace resolved from the authenticated bearer credential.
              required: true
              schema:
                $ref: '#/components/schemas/WorkspaceId'
        '400':
          description: Typed Checkfu wire error
          content:
            application/json:
              schema:
                type: object
                required:
                  - error
                properties:
                  error:
                    type: object
                    properties:
                      type:
                        type: string
                        enum:
                          - validation.malformed
                      message:
                        type: string
                      more:
                        type: string
                        enum:
                          - >-
                            https://docs.checkfu.com/reference/errors#validation-malformed
                        description: >-
                          Stable public documentation and remedy for this error
                          type.
                    required:
                      - type
                      - message
                      - more
                    additionalProperties: false
                additionalProperties: false
        '401':
          description: Typed Checkfu wire error
          content:
            application/json:
              schema:
                type: object
                required:
                  - error
                properties:
                  error:
                    type: object
                    required:
                      - type
                      - message
                      - more
                    properties:
                      type:
                        type: string
                        enum:
                          - auth.invalid_key
                      message:
                        type: string
                      more:
                        type: string
                        enum:
                          - >-
                            https://docs.checkfu.com/reference/errors#auth-invalid-key
                        description: >-
                          Stable public documentation and remedy for this error
                          type.
                    additionalProperties: false
                additionalProperties: false
        '403':
          description: >-
            The organization, tenant, or workspace backing this key is
            administratively disabled. | Deployment governance or retention
            policy denied the request.
          content:
            application/json:
              schema:
                anyOf:
                  - type: object
                    required:
                      - error
                    properties:
                      error:
                        type: object
                        properties:
                          type:
                            type: string
                            enum:
                              - auth.disabled_tenancy
                          message:
                            type: string
                          more:
                            type: string
                            enum:
                              - >-
                                https://docs.checkfu.com/reference/errors#auth-disabled-tenancy
                            description: >-
                              Stable public documentation and remedy for this
                              error type.
                        required:
                          - type
                          - message
                          - more
                        additionalProperties: false
                    additionalProperties: false
                  - $ref: '#/components/schemas/PolicyDeniedError'
        '404':
          description: >-
            The requested resource does not exist in the resolved deployment
            boundary.
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/ValidationNotFoundError'
        '409':
          description: >-
            The operation is illegal for the resource's current lifecycle state.
            | The request conflicts with the resource's current state. | An
            idempotent mutation conflicts with a completed or in-progress
            request for the same key.
          content:
            application/json:
              schema:
                anyOf:
                  - $ref: '#/components/schemas/RuntimeInvalidTransitionError'
                  - $ref: '#/components/schemas/ValidationConflictError'
                  - $ref: '#/components/schemas/IdempotencyConflictError'
        '422':
          description: No model candidate satisfies the deployment's routing constraints.
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/ModelNoEligibleModelError'
        '429':
          description: Typed Checkfu wire error
          headers:
            retry-after:
              description: >-
                Delay in seconds for rate limits or deployment quotas with a
                known release or UTC reset boundary
              required: false
              schema:
                type: integer
                minimum: 1
          content:
            application/json:
              schema:
                type: object
                required:
                  - error
                properties:
                  error:
                    type: object
                    required:
                      - type
                      - message
                      - more
                    properties:
                      type:
                        type: string
                        enum:
                          - budget.exceeded
                      message:
                        type: string
                      more:
                        type: string
                        enum:
                          - >-
                            https://docs.checkfu.com/reference/errors#budget-exceeded
                        description: >-
                          Stable public documentation and remedy for this error
                          type.
                    additionalProperties: false
                additionalProperties: false
        '500':
          description: >-
            An unexpected internal failure occurred; the message contains an
            opaque incident reference.
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/RuntimeInternalError'
      security:
        - bearerAuth: []
components:
  schemas:
    InstructionProposalId:
      type: string
      allOf:
        - pattern: ^inp_[0-9a-f]{32}$
    PrincipalId:
      type: string
      allOf:
        - pattern: ^prin_[0-9a-f]{32}$
    InstructionProposalConsistency_1:
      type: object
      properties:
        id:
          $ref: '#/components/schemas/InstructionProposalId'
        status:
          $ref: '#/components/schemas/Union_159'
        agent:
          $ref: '#/components/schemas/AgentId'
        base_draft_revision:
          type: integer
          allOf:
            - exclusiveMinimum: 0
        base_instructions:
          type: string
        model_routing_profile_key:
          type: string
          allOf:
            - pattern: ^[a-z0-9](?:[a-z0-9._-]{0,93})$
        judge_model_routing_profile_key:
          type: string
          allOf:
            - pattern: ^[a-z0-9](?:[a-z0-9._-]{0,93})$
        source_session_ids:
          $ref: '#/components/schemas/Arrays_32'
        instructions:
          anyOf:
            - type: string
              allOf:
                - maxLength: 4096
            - type: 'null'
        length_budget:
          type: integer
          allOf:
            - exclusiveMinimum: 0
        deltas:
          $ref: '#/components/schemas/Union_160'
        result_instructions:
          anyOf:
            - type: string
            - type: 'null'
        evaluation_material:
          $ref: '#/components/schemas/Union_161'
        evaluations:
          $ref: '#/components/schemas/Arrays_33'
        application:
          $ref: '#/components/schemas/Union_162'
        published_version:
          anyOf:
            - type: integer
              allOf:
                - exclusiveMinimum: 0
            - type: 'null'
        error:
          anyOf:
            - type: string
              enum:
                - input_session_unavailable
                - input_sessions_too_large
                - authoring_failed
                - internal
            - type: 'null'
        usage:
          $ref: '#/components/schemas/Objects_69'
        created_at:
          type: string
        updated_at:
          type: string
        version:
          type: integer
          allOf:
            - exclusiveMinimum: 0
      required:
        - id
        - status
        - agent
        - base_draft_revision
        - base_instructions
        - model_routing_profile_key
        - judge_model_routing_profile_key
        - source_session_ids
        - instructions
        - length_budget
        - deltas
        - result_instructions
        - evaluation_material
        - evaluations
        - application
        - published_version
        - error
        - usage
        - created_at
        - updated_at
        - version
      additionalProperties: false
    WorkspaceId:
      type: string
      allOf:
        - pattern: ^wrkspc_[0-9a-f]{32}$
    PolicyDeniedError:
      type: object
      properties:
        error:
          type: object
          properties:
            type:
              type: string
              enum:
                - policy.denied
            message:
              type: string
            more:
              type: string
              enum:
                - https://docs.checkfu.com/reference/errors#policy-denied
              description: Stable public documentation and remedy for this error type.
          required:
            - type
            - message
            - more
          additionalProperties: false
      required:
        - error
      additionalProperties: false
      description: Deployment governance or retention policy denied the request.
    ValidationNotFoundError:
      type: object
      properties:
        error:
          type: object
          properties:
            type:
              type: string
              enum:
                - validation.not_found
            message:
              type: string
            more:
              type: string
              enum:
                - https://docs.checkfu.com/reference/errors#validation-not-found
              description: Stable public documentation and remedy for this error type.
          required:
            - type
            - message
            - more
          additionalProperties: false
      required:
        - error
      additionalProperties: false
      description: >-
        The requested resource does not exist in the resolved deployment
        boundary.
    RuntimeInvalidTransitionError:
      type: object
      properties:
        error:
          type: object
          properties:
            type:
              type: string
              enum:
                - runtime.invalid_transition
            message:
              type: string
            more:
              type: string
              enum:
                - >-
                  https://docs.checkfu.com/reference/errors#runtime-invalid-transition
              description: Stable public documentation and remedy for this error type.
          required:
            - type
            - message
            - more
          additionalProperties: false
      required:
        - error
      additionalProperties: false
      description: The operation is illegal for the resource's current lifecycle state.
    ValidationConflictError:
      type: object
      properties:
        error:
          type: object
          properties:
            type:
              type: string
              enum:
                - validation.conflict
            message:
              type: string
            more:
              type: string
              enum:
                - https://docs.checkfu.com/reference/errors#validation-conflict
              description: Stable public documentation and remedy for this error type.
          required:
            - type
            - message
            - more
          additionalProperties: false
      required:
        - error
      additionalProperties: false
      description: The request conflicts with the resource's current state.
    IdempotencyConflictError:
      anyOf:
        - $ref: '#/components/schemas/ValidationIdempotencyConflictError'
        - $ref: '#/components/schemas/ValidationIdempotencyInProgressError'
      description: >-
        An idempotent mutation conflicts with a completed or in-progress request
        for the same key.
    ModelNoEligibleModelError:
      type: object
      properties:
        error:
          type: object
          properties:
            type:
              type: string
              enum:
                - model.no_eligible_model
            message:
              type: string
            more:
              type: string
              enum:
                - >-
                  https://docs.checkfu.com/reference/errors#model-no-eligible-model
              description: Stable public documentation and remedy for this error type.
          required:
            - type
            - message
            - more
          additionalProperties: false
      required:
        - error
      additionalProperties: false
      description: No model candidate satisfies the deployment's routing constraints.
    RuntimeInternalError:
      type: object
      properties:
        error:
          type: object
          properties:
            type:
              type: string
              enum:
                - runtime.internal
            message:
              type: string
            more:
              type: string
              enum:
                - https://docs.checkfu.com/reference/errors#runtime-internal
              description: Stable public documentation and remedy for this error type.
          required:
            - type
            - message
            - more
          additionalProperties: false
      required:
        - error
      additionalProperties: false
      description: >-
        An unexpected internal failure occurred; the message contains an opaque
        incident reference.
    Union_159:
      type: string
      enum:
        - authoring
        - drafted
        - authoring_failed
        - evaluating
        - evaluated
        - evaluation_failed
        - applied
        - discarded
    AgentId:
      type: string
      allOf:
        - pattern: ^agent_[0-9a-f]{32}$
    Arrays_32:
      type: array
      items:
        $ref: '#/components/schemas/SessionId'
      allOf:
        - minItems: 1
        - maxItems: 8
    Union_160:
      anyOf:
        - type: array
          items:
            anyOf:
              - type: object
                properties:
                  op:
                    type: string
                    enum:
                      - add
                  after:
                    anyOf:
                      - type: string
                        allOf:
                          - minLength: 1
                          - maxLength: 1024
                      - type: 'null'
                  text:
                    type: string
                    allOf:
                      - minLength: 1
                      - maxLength: 1024
                  provenance:
                    $ref: '#/components/schemas/Objects_88'
                required:
                  - op
                  - after
                  - text
                  - provenance
                additionalProperties: false
              - type: object
                properties:
                  op:
                    type: string
                    enum:
                      - replace
                  find:
                    type: string
                    allOf:
                      - minLength: 1
                      - maxLength: 1024
                  replace:
                    type: string
                    allOf:
                      - minLength: 1
                      - maxLength: 1024
                  provenance:
                    $ref: '#/components/schemas/Objects_88'
                required:
                  - op
                  - find
                  - replace
                  - provenance
                additionalProperties: false
              - type: object
                properties:
                  op:
                    type: string
                    enum:
                      - retire
                  find:
                    type: string
                    allOf:
                      - minLength: 1
                      - maxLength: 1024
                  provenance:
                    $ref: '#/components/schemas/Objects_88'
                required:
                  - op
                  - find
                  - provenance
                additionalProperties: false
          allOf:
            - minItems: 1
            - maxItems: 8
        - type: 'null'
    Union_161:
      anyOf:
        - type: object
          properties:
            task:
              type: string
              allOf:
                - minLength: 1
                - maxLength: 1024
            rubric:
              type: string
              allOf:
                - minLength: 1
                - maxLength: 2048
          required:
            - task
            - rubric
          additionalProperties: false
        - type: 'null'
    Arrays_33:
      type: array
      items:
        $ref: '#/components/schemas/InstructionEvaluationDistinctSessions'
      allOf:
        - maxItems: 4
    Union_162:
      anyOf:
        - type: object
          properties:
            agent_version:
              type: integer
              allOf:
                - exclusiveMinimum: 0
            draft_revision:
              type: integer
              allOf:
                - exclusiveMinimum: 0
            applied_at:
              type: string
          required:
            - agent_version
            - draft_revision
            - applied_at
          additionalProperties: false
        - type: 'null'
    Objects_69:
      type: object
      properties:
        input_tokens:
          type: integer
          allOf:
            - minimum: 0
        cached_input_tokens:
          type: integer
          allOf:
            - minimum: 0
        cache_write_input_tokens:
          type: integer
          allOf:
            - minimum: 0
        output_tokens:
          type: integer
          allOf:
            - minimum: 0
        reasoning_output_tokens:
          type: integer
          allOf:
            - minimum: 0
        total_tokens:
          type: integer
          allOf:
            - minimum: 0
      required:
        - input_tokens
        - cached_input_tokens
        - cache_write_input_tokens
        - output_tokens
        - reasoning_output_tokens
        - total_tokens
      additionalProperties: false
    ValidationIdempotencyConflictError:
      type: object
      properties:
        error:
          type: object
          properties:
            type:
              type: string
              enum:
                - validation.idempotency_conflict
            message:
              type: string
            more:
              type: string
              enum:
                - >-
                  https://docs.checkfu.com/reference/errors#validation-idempotency-conflict
              description: Stable public documentation and remedy for this error type.
          required:
            - type
            - message
            - more
          additionalProperties: false
      required:
        - error
      additionalProperties: false
      description: The Idempotency-Key is already bound to a different request.
    ValidationIdempotencyInProgressError:
      type: object
      properties:
        error:
          type: object
          properties:
            type:
              type: string
              enum:
                - validation.idempotency_in_progress
            message:
              type: string
            more:
              type: string
              enum:
                - >-
                  https://docs.checkfu.com/reference/errors#validation-idempotency-in-progress
              description: Stable public documentation and remedy for this error type.
          required:
            - type
            - message
            - more
          additionalProperties: false
      required:
        - error
      additionalProperties: false
      description: >-
        An identical idempotent request is still in progress and may be retried
        later.
    SessionId:
      type: string
      allOf:
        - pattern: ^sess_[0-9a-f]{32}$
    Objects_88:
      type: object
      properties:
        sources:
          type: array
          items:
            type: object
            properties:
              session:
                $ref: '#/components/schemas/SessionId'
              event:
                $ref: '#/components/schemas/EventId'
            required:
              - session
              - event
            additionalProperties: false
          allOf:
            - maxItems: 8
        note:
          anyOf:
            - type: string
              allOf:
                - maxLength: 512
            - type: 'null'
      required:
        - sources
        - note
      additionalProperties: false
    InstructionEvaluationDistinctSessions:
      type: object
      properties:
        baseline_session:
          $ref: '#/components/schemas/SessionId'
        candidate_session:
          $ref: '#/components/schemas/SessionId'
        started_at:
          type: string
        result:
          anyOf:
            - type: object
              properties:
                baseline:
                  $ref: '#/components/schemas/Union_155'
                candidate:
                  $ref: '#/components/schemas/Union_155'
                improved:
                  type: boolean
                baseline_cost_tokens:
                  anyOf:
                    - type: integer
                      allOf:
                        - minimum: 0
                    - type: 'null'
                candidate_cost_tokens:
                  anyOf:
                    - type: integer
                      allOf:
                        - minimum: 0
                    - type: 'null'
                completed_at:
                  type: string
              required:
                - baseline
                - candidate
                - improved
                - baseline_cost_tokens
                - candidate_cost_tokens
                - completed_at
              additionalProperties: false
            - type: 'null'
      required:
        - baseline_session
        - candidate_session
        - started_at
        - result
      additionalProperties: false
    EventId:
      type: string
      allOf:
        - pattern: ^evt_[0-9a-f]{32}$
    Union_155:
      type: string
      enum:
        - satisfied
        - needs_revision
        - max_iterations_reached
        - failed
        - interrupted
  securitySchemes:
    bearerAuth:
      type: http
      scheme: bearer

````