No. Anthropic Claude API returns a success response when it accepts the request. The message_batch then holds one of 6 states. Only succeeded is terminal success. Provely proves all_succeeded and ended as separate promises.

Skill facts from the signed manifest.
FactValue
Skill version0.1.0
PublisherBuilt and signed by Provely.
CertificationProvisional (score 49 of 100)
Last conformance run2026-09-05T12:00:00Z: 26 of 26 cases passed, 0 critical false VERIFIED
Provider API versions2023-06-01
Default provider API version2023-06-01
Compiled2026-09-05T12:00:00Z by compiler 0.1.0
Manifest hash1570330057d84a4b
Manifest hash checkthe document hashes to the value the manifest states
Signaturevalid, key provely-skill-2026-09, trusted by this build

What does this page prove?

ClaimProven byEvidenceStatus
A batch that ended is not a batch whose requests succeeded. The level reads request_counts and refuses a batch that holds an error, an expiry, or a cancellation.anthropic.message_batch.all_succeededE2proven
Every request of the batch succeeded, errored, was canceled, or expired. The level proves that the work stopped. It proves no result of any request.anthropic.message_batch.endedE2proven
An outcome outside Anthropic Claude API, such as a bank credit or a person who read a messagenot provenno E5 channelnot proven
The agent report that the action workednever countsE0not proven

Which completion levels does the Anthropic Claude API skill expose?

Each level is one promise with one contract. An agent picks the level that matches the promise it makes. It cannot upgrade a level. Read the completion level definition.

LevelContractPromiseEvidenceCertification
all_succeededanthropic.message_batch.all_succeeded v1.0.0A batch that ended is not a batch whose requests succeeded. The level reads request_counts and refuses a batch that holds an error, an expiry, or a cancellation.E2Provisional
endedanthropic.message_batch.ended v1.0.0Every request of the batch succeeded, errored, was canceled, or expired. The level proves that the work stopped. It proves no result of any request.E2Provisional

What is the Anthropic Claude API lifecycle?

Which states can a anthropic.message_batch be in?

StateClassVerdictMeaningSource
in_progresstransitionalPENDINGAnthropic accepted the batch. Every terminal count of request_counts is zero, so the read reports no success and no failure.anthropic.docs.batches
cancelingtransitionalPENDINGA caller asked for a cancellation. Anthropic may still complete a request that it cannot interrupt.anthropic.docs.batches
succeededterminal successVERIFIEDThe member processing_status is ended, and request_counts.succeeded holds every request of the batch.anthropic.docs.batches
erroredterminal failureFAILEDThe member processing_status is ended, and request_counts.errored is above zero. Anthropic created no message for those requests.anthropic.docs.batches
expiredterminal failureFAILEDThe batch reached the 24-hour expiration, and request_counts.expired is above zero. Anthropic sent those requests to no model.anthropic.docs.batches
canceledterminal neutralFAILEDA caller canceled the batch, and request_counts.canceled is above zero. The batch still reports the status ended.anthropic.docs.batches

In anthropic.message_batch under provider API version 2023-06-01, succeeded is the only state that means terminal success. Every other state gives PENDING, FAILED, or UNVERIFIABLE.

Source: anthropic.docs.batches · retrieved 2026-09-09

How does Provely tie the evidence to this exact operation?

A matching state that already existed must not verify. Every contract names the correlation keys that bind the evidence to the operation, and the idempotency key that stops a duplicate side effect.

StrategyAssuranceKeysRequiredWindow
resource_idstrongmessage_batch_id from $action.result.idyesnone
The skill reads the channels below. It prefers the ones furthest from the action.
In words
  • E0 agent assertion: never sufficient.
  • E1 action response: the provider acknowledged the request.
  • E2 provider readback: the runtime read the resource back.
  • E3 provider event: the provider reported the change.
  • E4 independent system: a system outside the action path agrees.
  • E5 external outcome: the result is observable in the world.

Which evidence channels does the skill read?

The runtime prefers the channel that is more independent from the action path. Read the evidence level definition. An acknowledgement from Anthropic Claude API is E1 and never terminal success.

ChannelLevelIndependenceVerifierDeterministicTypical latency
message_batch_action_responseE1same responseaction_resultyesnot stated
message_batch_readbackE2provider readbackhttpyes300 ms

Which ways can a Anthropic Claude API action look done and not be?

ContractCaseRuleVerdict
anthropic.message_batch.all_succeededpre existing stateThe Message Batch is older than the operation. It proves nothing.CONTRADICTED
anthropic.message_batch.all_succeededwrong subjectThe batch completed more requests than the intent named. It is not the batch that the intent described.CONTRADICTED
anthropic.message_batch.all_succeededduplicate side effectThe batch holds more requests than the intent named. Anthropic bills each of them. Do not retry.CONTRADICTED
anthropic.message_batch.all_succeededobserved stateA caller canceled the batch. Anthropic sent those requests to no model.CONTRADICTED
anthropic.message_batch.all_succeededobserved stateThe batch reached the 24-hour expiration before Anthropic sent every request to a model.FAILED
anthropic.message_batch.all_succeededobserved stateA request of the batch encountered an error, and Anthropic created no message.FAILED
anthropic.message_batch.all_succeededobserved stateThe Message Batch has not ended. The runtime observes again later.PENDING
anthropic.message_batch.endedpre existing stateThe Message Batch is older than the operation. It proves nothing.CONTRADICTED
anthropic.message_batch.endedwrong subjectThe batch completed more requests than the intent named. It is not the batch that the intent described.CONTRADICTED
anthropic.message_batch.endedduplicate side effectThe batch holds more requests than the intent named. Anthropic bills each of them. Do not retry.CONTRADICTED
anthropic.message_batch.endedobserved stateThe Message Batch has not ended. The runtime observes again later.PENDING

What did the last conformance run show?

The six confidence dimensions of the signed manifest.
DimensionScoreMaximum
Documentation1220
Schema alignment1515
Lifecycle certainty2020
Evidence strength1220
Causal correlation1515
Sandbox conformance1010
Raw score84100
Score after the caps49100

The raw score is 84. 2 hard caps apply, so the score is 49 and the level is Provisional.

Why the score is capped.
ReasonHighest score it allows
10 uncertainty records are open. The lowest cap is 49.49
1 mandatory cases are missing. Example: anthropic.message_batch.ended:terminal_failure.74
The score and the level of each contract.
ContractRaw scoreScore after the capsCertification
anthropic.message_batch.all_succeeded8449Provisional
anthropic.message_batch.ended8449Provisional

The weakest contract is anthropic.message_batch.all_succeeded. It scores 49 and reaches Provisional. The package level is Provisional, because a package level never rises above its weakest contract. Read the level of the contract you use.

The last conformance run on 2026-09-05T12:00:00Z passed 26 of 26 cases with 0 critical false VERIFIED. One critical false VERIFIED rejects a skill.

The conformance result by case class.
Case classPassedTotal
duplicate side effect22
error after execution22
error before execution22
evidence unavailable22
pre existing state22
stale readback22
still transitional22
terminal failure22
terminal success33
timeout after commit22
version mismatch22
wrong subject22
wrong terminal state11

The conformance artifact digest is eae45ca3048c69c3. The harness signs the run, so a reader can check that these numbers come from that run.

What remains uncertain?

The level is Provisional because of it. 10 uncertainty records are open. The lowest cap is 49. Provely does not guess a rule that a source does not state.

Where do these facts come from?

Every claim above cites a source assertion in the skill provenance. The compiler records the source, its hash, and the retrieval date. A page never states a provider rule without one.

SourceKindRetrievedExcerpt
anthropic.docs.batch_objectdocs2026-09-09authored
anthropic.docs.batchesdocs2026-09-09authored
anthropic.docs.filesdocs2026-09-09authored
anthropic.docs.versionsdocs2026-09-09authored
anthropic.openapiopenapi2026-09-09authored

How do I verify a Anthropic Claude API action?

Verify a Anthropic Claude API action with Provely

  1. Begin the operation.Call begin with the contract anthropic.message_batch.all_succeeded and the input. Keep the operation id.
  2. Make the Anthropic Claude API call you make today.Send the request with the correlation metadata that begin returned.
  3. Submit the acknowledgement.Call action_result with the Anthropic Claude API response. This is evidence level E1. It is not completion.
  4. Verify.Call verify. The runtime reads message_batch_action_response and message_batch_readback and evaluates the contract.
  5. Report the verdict exactly as returned.VERIFIED comes with a signed receipt. PENDING comes with the operation id. CONTRADICTED and UNVERIFIABLE are not success.

Questions developers ask

Does the all_succeeded level prove the ended level?

No. A batch that ended is not a batch whose requests succeeded. The level reads request_counts and refuses a batch that holds an error, an expiry, or a cancellation. Every request of the batch succeeded, errored, was canceled, or expired. The level proves that the work stopped. It proves no result of any request. Use anthropic.message_batch.ended to prove ended.

What does the agent say while Anthropic Claude API reports in_progress and canceling?

It says: "The action is accepted but not yet verified. Operation: <id>." The verdict is PENDING. The runtime observes again on the contract timing policy.

Which Anthropic Claude API API versions does the skill support?

2023-06-01. An operation on another version returns UNVERIFIABLE with the reason version_unsupported. The runtime never guesses.

Does Provely need write access to Anthropic Claude API?

No. The agent keeps its write key. The verifier reads with a separate read-only credential where Anthropic Claude API permits it, and it never shares that credential with the agent.