Enum AgentStyle 

pub enum AgentStyle {
    GeneralPurpose,
    Plan,
    Verification,
    Explore,
    CodeReview,
}

Agent style — determines which system prompt template is used.

Each style has a different focus and behavior, selected based on the user’s apparent intent from their message.

Variants

GeneralPurpose

Default — general purpose coding agent for research and multi-step tasks.

Plan

Read-only planning and architecture analysis. Prohibited from modifying files, focuses on design and planning.

Verification

Adversarial verification specialist — tries to break code, not confirm it works.

Explore

Fast file search and codebase exploration. Read-only, optimized for finding files and patterns quickly.

CodeReview

Code review focused — analyzes code quality, best practices, potential issues.

Implementations

impl AgentStyle

pub fn base_prompt(&self) -> &'static str

Returns the base system prompt for this style.

pub fn guidelines(&self) -> Option<&'static str>

Returns style-specific guidelines if any.

pub fn description(&self) -> &'static str

Returns a one-line description of this style.

pub fn requires_planning(&self) -> bool

Returns true if this style benefits from a planning phase.

Planning is beneficial for styles that involve multi-step execution or where a structured approach improves outcomes.

pub fn detect_with_confidence(message: &str) -> (Self, DetectionConfidence)

Detects the most appropriate agent style based on user message content, along with a confidence level.

Use this for fast-path keyword matching. When confidence is [Low], consider using detect_with_llm for more accurate classification.

pub fn detect_from_message(message: &str) -> Self

Detects the most appropriate agent style based on user message content.

This is a simple keyword-based heuristic. For more precise control, users can explicitly set the style via SystemPromptSlots::with_style().

pub async fn detect_with_llm(llm: &dyn LlmClient, message: &str) -> Result<Self>

Classifies user intent using LLM when keyword confidence is low.

This is called when [detect_with_confidence] returns [Low] confidence, indicating the message doesn’t have clear keyword indicators.

Uses a lightweight classification prompt that returns a single word.

Trait Implementations

impl Clone for AgentStyle

fn clone(&self) -> AgentStyle

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for AgentStyle

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

Formats the value using the given formatter.

impl Default for AgentStyle

fn default() -> AgentStyle

Returns the “default value” for a type.

impl PartialEq for AgentStyle

fn eq(&self, other: &AgentStyle) -> bool

Tests for self and other values to be equal, and is used by ==.

fn ne(&self, other: &Rhs) -> bool

Tests for !=. The default implementation is almost always sufficient, and should not be overridden without very good reason.

impl Copy for AgentStyle

impl Eq for AgentStyle

impl StructuralPartialEq for AgentStyle