← ClaudeAtlas

zig-agentslisted

Patterns and best practices for building AI agents in Zig. Covers tool systems, context management, LLM providers, streaming responses, and session persistence. Use when implementing agent functionality.
satibot/satibot · ★ 3 · AI & Automation · score 61
Install: claude install-skill satibot/satibot
# Zig Agent Development ## Overview This skill covers patterns for building AI agents in Zig, including: - Tool system design and implementation - Conversation context management - LLM provider abstractions - Streaming response handling - Session persistence - Agent loop patterns ## Architecture ```text Agent ├── Config # API keys, model settings ├── Context # Conversation history ├── ToolRegistry # Available tools ├── Provider # LLM API abstraction └── Session # Persistence layer ``` ### Agent Loop Pattern The core agent loop follows this flow: ```text 1. Add user message to context 2. Loop (max N iterations): a. Call LLM with context + tools b. Add assistant response to context c. If tool calls present: - Execute each tool - Add tool results to context - Continue loop d. Else: break 3. Save session ``` ## Tool System ### Tool Definition ```zig pub const ToolContext = struct { allocator: std.mem.Allocator, config: Config, }; pub const Tool = struct { name: []const u8, description: []const u8, parameters: []const u8, // JSON Schema execute: *const fn (ctx: ToolContext, arguments: []const u8) anyerror![]const u8, }; ``` ### Tool Registry ```zig pub const ToolRegistry = struct { tools: std.StringHashMap(Tool), allocator: std.mem.Allocator, pub fn init(allocator: std.mem.Allocator) ToolRegistry { return .{ .tools = std.Strin