Enum Decision 

pub enum Decision {
    Allow,
    Deny(String),
    Blocked(String),
}

The decision an approver hands back for one pending call.

Two ways to say no, and the difference is load-bearing rather than cosmetic. The learning miner keys on the exact string "Denied by the user:" to find corrections worth learning from, so a refusal that no human made must not wear that label. Blocked is the machine’s no — a permission mode, a policy, a remote prompt nobody answered — and the loop renders it as "Blocked by policy:", joining "Blocked by a hook:" in the family of refusals the miner ignores.

Without the split, a read-only run’s refusals and a 2am approval nobody was awake to answer both become training data attributed to a user who never spoke. It is the same mistake as mining a publish’s changed path as a voice correction, and it was live in ModeApprover until a Slack approver needed to express “nobody answered” and found there was no way to.

Variants

Allow

Deny(String)

A human said no. The reason is passed to the model so it can pick another approach — and it is mined as a correction.

Blocked(String)

Machine policy said no, and no human was consulted. Never mined.

Trait Implementations

impl Clone for Decision

fn clone(&self) -> Decision

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for Decision

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.