See exactly what the review delivers.
This fictional report shows how Turncoat turns one agent configuration into a clear rollout decision: what you intended, what the evidence shows, and what still needs proof. It is not a customer result or a measured assessment.
Start with a pinned decision
Agent: Example Agent · Version: 1.2.3 · Configuration: interactive, approvals enabled · Snapshot: illustrative only; no executable artifact was assessed.
Decision: whether workspace writes and network access match the organization’s intended rules. A real review pins the version, configuration, and evidence before drawing a conclusion.
See intent beside evidence
| Capability | Intent | Example observation | Review |
|---|---|---|---|
| Workspace writes | Ask | Boundary held | Consistent within the example |
| Network access | Deny | Not established | Evidence needed |
| Read designated project files | Allow | Executes without asking | Intentionally allowed behavior |
The table is the point: intended behavior sits beside the observed result and the review status. An untested capability stays Not established; it never becomes an implicit allow.
The evidence bar
A report distinguishes an attempted action, a recorded execution, and a confirmed effect. It records the source, matched control outcome, normal-task completion, and incomplete observations.
This sample asserts no execution or effect. In a real review, an error or a control that failed its normal task cannot establish a clean contrast.
Turn the sample into your decision
For this fictional example, network access remains unresolved. A real rollout owner would need evidence for that capability before treating it as covered. Changes to the version, snapshot, or flags require a review of the affected conclusions.
The deployer review covers recorded, controlled evidence and configuration. Findings remain private, and no outcome is guaranteed.