Crate ferrify_domain 

Core domain types for Ferrify.

agent-domain is the vocabulary crate for the rest of the workspace. It defines the value objects, planning records, policy types, provenance labels, and reporting structures that let Ferrify describe a governed software-change run without reaching into filesystem or process concerns.

The main design goal is to keep meaning-bearing concepts explicit. A path that must stay inside the repository is represented by RepoPath, not a raw String. A mode identifier is a validated ModeSlug, not a free-form label. The result is a control plane that can encode authority and scope in the type system before any command is run.

Core Concepts

• PolicyLayer, TrustLevel, and Capability describe who may do what, and why.
• ChangeIntent, ChangePlan, and PatchPlan carry work from intake into a bounded implementation strategy.
• InputRole and ClassifiedInput explain how Ferrify separates operator goals, repository policy, code, evidence, and untrusted text.
• FinalChangeReport and ValidationReceipt make reporting evidence-backed instead of speculative.

Examples

use agent_domain::{ModeSlug, RepoPath, TrustLevel};

let target = RepoPath::new("crates/agent-cli/src/main.rs")?;
let mode = ModeSlug::new("architect")?;

assert_eq!(target.as_str(), "crates/agent-cli/src/main.rs");
assert_eq!(mode.as_str(), "architect");
assert!(TrustLevel::RepoPolicy.can_define_policy());

Structs

ApprovalProfileSlug

A stable slug identifying an approval profile.

ArtifactRef

A reference to evidence captured during a verification step.

Assumption

An assumption that influenced the run.

ChangeIntent

The intake object that describes the requested software change.

ChangePlan

The architect-stage change plan.

ChangeSummary

The high-level change outcome shown to the operator.

ClassifiedInput

A lightweight provenance record for an input observed during a run.

EffectivePolicy

The fully resolved policy used by a run.

EvidenceRequirement

A verification requirement requested by the operator or planner.

FinalChangeReport

The evidence-backed final report.

ModeSlug

A stable slug identifying an execution mode.

OutcomeSpec

The desired outcome stated for the task.

PatchAnchor

A file-level anchor used by patch planning.

PatchBudget

The patch budget used to keep edits bounded.

PatchPlan

The implementer-stage patch plan.

PathPattern

A repository-relative path restriction.

RepoPath

A repository-relative path that cannot escape the workspace root.

ReportingPolicy

Reporting rules that constrain what the runtime may claim.

RiskItem

A risk item preserved across planning and reporting.

ScopeBoundary

The explicit scope boundary for a task.

ScopeItem

A repository item used to bound scope.

TouchedArea

A touched area that should appear in the final report.

ValidationMinimums

Minimum verification requirements imposed by policy.

ValidationReceipt

A receipt proving which verification command ran and how it ended.

VerificationPlan

The verification steps required for a plan.

Enums

ApiImpact

The expected API impact for the change.

ApprovalRule

The approval policy attached to a capability.

BlastRadius

The maximum acceptable blast radius for a change.

Capability

A capability that can be granted, denied, or approval-gated.

ChangeStatus

The overall status for the final report.

DependencyPolicy

The repository stance on dependency changes.

DomainTypeError

Validation failures for domain value objects.

InputRole

The operational role an input plays in Ferrify.

PolicyLayer

The precedence layer that introduced a policy fragment.

RiskLevel

The severity attached to a residual risk.

SemanticConcern

The semantic concern for a patch.

TaskKind

The kind of change the operator is requesting.

TrustLevel

The trust classification attached to inputs and observations.

VerificationKind

The verification steps supported by Ferrify.

VerificationStatus

The result of an individual verification step.