Enum ToolVerdict 

#[non_exhaustive]
pub enum ToolVerdict {
    Allowed {
        tool: String,
        budget_remaining: u64,
        elapsed_us: u64,
        payout_micros: u64,
        principal_id: Option<String>,
    },
    Blocked {
        tool: String,
        stage: &'static str,
        policy: Option<String>,
        denial_code: DenialCode,
        reason: String,
        response: Value,
        elapsed_us: u64,
    },
}

Verdict for one intercepted tool call.

Variants (Non-exhaustive)

Non-exhaustive enums could have additional variants added in future. Therefore, when matching against variants of non-exhaustive enums, an extra wildcard arm must be added to account for any future variants.

Allowed

Forward to the downstream tool server.

Fields

tool: String

Tool name.

budget_remaining: u64

Remaining headroom under the binding budget dimension.

elapsed_us: u64

Decision latency in microseconds (SLA surface, R23).

payout_micros: u64

Payout amount (USD micros) that ActionBudget::try_tool_call debited on this call. Threaded out so a lost execution.claim() race in the harness can call [ActionBudget::refund_tool_call] with the exact same amount, closing the concurrent-MCP budget double-spend.

principal_id: Option<String>

Principal id whose ledger was ALSO debited (when a principal-scoped budget is bound). Threaded out so every harness failure path that refunds the session ledger can mirror the refund on the principal ledger — without this, each lost claim race / saturated worker permanently consumed principal quota with zero tools executed.

Blocked

Block; respond to the agent with response.

Fields

tool: String

Tool name (<unparsed> when the payload never parsed).

stage: &'static str

Which gate blocked: parse | schema | policy | budget.

policy: Option<String>

Name of the policy engine that denied, when the block came from the named policy chain (stage: "policy"). Parse, schema, and budget gates are configured limits rather than named policies, so they stay None — the console’s per-policy attribution must never invent a policy that the operator can’t find in their policy inventory.

denial_code: DenialCode

Structured denial code for machine consumption.

reason: String

Human/machine-readable reason.

response: Value

Ready-to-send JSON-RPC authorization error.

elapsed_us: u64

Decision latency in microseconds.

Implementations

impl ToolVerdict

pub fn is_allowed(&self) -> bool

True when the call may proceed.

pub fn elapsed_us(&self) -> u64

Decision latency in microseconds.

Trait Implementations

impl Debug for ToolVerdict

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

Formats the value using the given formatter.