Struct ActionPolicy 

pub struct ActionPolicy {
    pub id: String,
    pub name: String,
    pub mode: EnforcementMode,
    pub allowed_tools: Option<HashSet<String>>,
    pub blocked_tools: HashSet<String>,
    pub max_risk_score: f64,
    pub allowed_action_types: Option<HashSet<AgentActionType>>,
    pub max_actions_per_session: Option<u32>,
    pub allow_unregistered: bool,
}

Policy for controlling which actions an agent can take.

Combines allowlist/blocklist enforcement, risk score thresholds, action type filtering, and session-level rate limiting into a single evaluable policy.

Use the builder methods to configure, or the convenience constructors ActionPolicy::permissive and ActionPolicy::restrictive.

Fields

id: String

Policy identifier.

name: String

Human-readable name.

mode: EnforcementMode

Enforcement mode.

allowed_tools: Option<HashSet<String>>

Allowed tool IDs (if set, only these tools are permitted).

blocked_tools: HashSet<String>

Blocked tool IDs (these tools are always denied).

max_risk_score: f64

Maximum risk score allowed (tools with higher risk are blocked).

allowed_action_types: Option<HashSet<AgentActionType>>

Allowed action types (if set, only these types are permitted).

max_actions_per_session: Option<u32>

Maximum total actions per session.

allow_unregistered: bool

Whether to allow actions on unregistered tools.

Implementations

impl ActionPolicy

pub fn new(id: &str, name: &str) -> Self

Create a new action policy with sensible defaults.

Defaults: enforce mode, no allowlist, no blocklist, max risk 1.0, all action types allowed, no session limit, unregistered tools allowed.

pub fn permissive(id: &str, name: &str) -> Self

Create a permissive policy that allows everything in audit mode.

All actions are allowed; violations are only logged as findings.

pub fn restrictive(id: &str, name: &str) -> Self

Create a restrictive policy that denies by default.

Requires explicit allowlist, blocks unregistered tools, and enforces a conservative risk threshold of 0.7.

pub fn with_mode(self, mode: EnforcementMode) -> Self

Set the enforcement mode.

pub fn with_allowed_tools(self, tools: HashSet<String>) -> Self

Set the allowed tool IDs. Only these tools will be permitted.

pub fn with_blocked_tools(self, tools: HashSet<String>) -> Self

Set the blocked tool IDs. These tools are always denied.

pub fn with_max_risk_score(self, score: f64) -> Self

Set the maximum risk score allowed.

pub fn with_allowed_action_types(self, types: HashSet<AgentActionType>) -> Self

Set the allowed action types.

pub fn with_max_actions_per_session(self, max: u32) -> Self

Set the maximum actions per session.

pub fn with_allow_unregistered(self, allow: bool) -> Self

Set whether unregistered tools are allowed.

pub fn evaluate( &self, action: &AgentAction, tool_def: Option<&ToolDefinition>, ) -> PolicyDecision

Evaluate an action against this policy.

Returns a PolicyDecision indicating whether the action is allowed, denied, or warned, along with any SecurityFindings.

Trait Implementations

impl Clone for ActionPolicy

fn clone(&self) -> ActionPolicy

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for ActionPolicy

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

Formats the value using the given formatter.