mirror of
https://github.com/j3ssie/osmedeus.git
synced 2026-09-22 09:34:57 +02:00
- Add AgentExecutor implementing LLM-based agentic loop with tool calling, max iterations, and stop conditions - Introduce agent preset tools (bash, file_exists, http_get, run_module, etc.) with extensible registry pattern - Add sub-agent spawning capability via spawn_agent tool call with recursive depth limits and validation - Implement ToolExecutor for custom tool execution with template rendering and error handling - Add agent session persistence and memory management with sliding window configuration - Create comprehensive E2E test suite covering 15+ agent workflow scenarios (minimal, custom tools, planning, multi-goal, structured output, tracing hooks, file tools, orchestration, Python tools, sub-agents, nested sub-agents, and validation) - Add agent-and-llm test data directory with 17 YAML workflow fixtures - Update integration tests to include agent workflow directories - Add AgentTool and AgentConfig types with validation for duplicate names and unknown presets - Implement LLM streaming test utilities - Update documentation (CLAUDE.md, HACKING.md, README.md) with agent features and CLI examples
29 lines
701 B
YAML
29 lines
701 B
YAML
kind: module
|
|
name: test-agent-sub-agents-validation-fail
|
|
description: Agent with duplicate sub-agent names (should fail validation)
|
|
tags: test,agent,validation-fail
|
|
|
|
params:
|
|
- name: target
|
|
required: true
|
|
default: example.com
|
|
|
|
steps:
|
|
- name: bad-orchestrator
|
|
type: agent
|
|
query: "Analyze {{Target}}"
|
|
max_iterations: 5
|
|
agent_tools:
|
|
- preset: bash
|
|
sub_agents:
|
|
- name: recon_agent
|
|
description: "First recon agent"
|
|
max_iterations: 3
|
|
agent_tools:
|
|
- preset: bash
|
|
- name: recon_agent
|
|
description: "Duplicate name - should cause validation error"
|
|
max_iterations: 3
|
|
agent_tools:
|
|
- preset: bash
|