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

# Verified work

> Read destination-backed outcomes, human-review readiness, and scoped change comparisons in Tensile.

Verified work answers three separate questions about an agent workflow. These views appear in an organization once the relevant task contract and evidence have been configured.

| View          | Question                                                                     | What to inspect                                                                                             |
| ------------- | ---------------------------------------------------------------------------- | ----------------------------------------------------------------------------------------------------------- |
| **Outcomes**  | Did the requested work happen at the destination?                            | The declared task, the current case verdict, destination observations, and any linked execution.            |
| **Readiness** | Is the evidence strong enough to change the human-review level?              | The frozen cohort, coverage and qualification conditions, then the separately recorded owner choice.        |
| **Changes**   | Did a candidate improve a frozen set of cases without a declared regression? | The baseline and candidate evidence for each case, the scoped comparison result, and any owner disposition. |

An agent's success claim or a tool acknowledgment describes what the agent reported. It does not by itself prove that a message was sent to the intended channel and thread, or that any other destination changed. An indexed execution helps explain the agent's activity; an Outcome case checks the declared work against destination evidence.

## Follow a case

1. Open **Outcomes** and select a business case. The top of the case file shows the current verdict and the difference between what the task allowed and what the destination recorded.
2. Open **Destination evidence** to inspect each new record, its destination, timestamp, and text-fingerprint match. The full source references and assessment revisions are available in the audit details. If evidence has expired, become unavailable, or been superseded, the page distinguishes the historical result from current proof.
3. Follow **View indexed agent run** when an execution is linked. The run identity and chronology are supporting context; the destination records determine the business outcome.
4. Open **Readiness** to see whether the case's frozen cohort supports a change in human review. The decision and blockers come first; the policy, assessment checks, and segment table can be expanded.
5. Open **Changes** to compare a candidate with a baseline on the same declared cases. A result that passes its test scope is limited to those cases. It is not permission to deploy.

Readiness records an assessment and an owner decision separately. Recording a decision in Tensile does not switch live review routing. Missing or unqualified source evidence cannot be turned into a passing review-reduction claim by excluding it from the table.

## Try the Tensile example

If your account has access, open the organization selector at the top of the left sidebar and choose **Tensile Outcome Demo**. On a narrow screen, open navigation from the top-left control first. Then open **Outcomes → Duplicate**.

The authored Slack task allowed **one** new reply in a specified workspace, channel, and thread. The agent reported success, but the destination fixture contains **two** matching new replies, so the Outcome is **Failed**. In **Readiness**, the seven-case authored cohort remains on full review because five evidence conditions are unmet. From the Duplicate case, **Open baseline and candidate evidence** shows a separately authored Change comparison: the baseline has two writes, the candidate has one, and the saved result passes only that frozen test scope.

<Note>
  This organization is an authored demonstration. Its trace events were ingested and indexed, while the Slack destination observations and comparison arms were prepared separately. The example did not send or read back a live Slack message, rerun an agent in a clone, obtain customer approval, or authorize a release.
</Note>

## Use evidence from your own workflow

A real-source Outcome needs a confirmed task expectation, an authorized destination source that can provide the required read-back, and attributable case identity. Readiness additionally needs a frozen cohort and validated checks before an owner can consider reduced review. A Change comparison needs independently pinned baseline and candidate evidence for its declared cases. If a view is absent for your organization, those records or capabilities have not been configured there; a trace connection alone does not enable verified work.