# Outcomes

*The delivered work product of a run.*

An outcome is what you actually get from a run: a reviewable, approvable work
product with full provenance.

## The Proof Rail

Every outcome carries a **Proof Rail** — an evidence snapshot with source
versions, the method and version used, the runtime / model / tool policy, cost,
and the full approval trail. Approved publications are recorded as immutable,
versioned **gate results**, and sign-offs produce **signed deliverables** with
maker–checker signatures and a content hash.

## Needs You

Outcomes separate the machine's work from the **true human decisions**. Anything
that genuinely needs a person shows up in **Needs You**, so review is about
decisions, not digging.

## Reconciliation

Recurring services fold their runs into **one living outcome** — updated in place
with a version trail — so you see what's new, open, and resolved over time
instead of a stack of separate reports.

## Connected work

An outcome or its artifacts can be **handed off** to a downstream service,
linking work across runs so provenance follows the chain.

## Approving & exporting

Review happens on the outcome itself. Depending on the service you can **request
review**, **approve**, apply a **signature**, and **export** — and these
side-effecting actions run through governed steps, so sensitive links and exports
are opened via a controlled action rather than exposed directly. Sign-offs produce
signed deliverables with maker–checker signatures and a content hash.
