How does it prevent agents from overwriting each other's work?

It enforces strict file ownership, assigning unique, non-overlapping file lists to each agent during the dispatch phase to prevent write conflicts.

Can I trigger the Agent Dispatcher manually?

While it can be triggered by specific phrases like 'run agents', it is designed as a non-interactive skill that typically follows the planning and decomposition phase.

What are reference paths in this context?

Reference paths are project-specific documentation or knowledge files identified in your project map that the dispatcher provides to agents to ensure they have proper context for their tasks.

Does Agent Dispatcher handle task planning?

No, it works downstream from the task-decomposer and tier-assignment skills to execute pre-defined plans and assignments.

What happens if a subtask fails?

The skill automatically retries the task once at the same model tier. If it fails a second time, it escalates the task to a higher-tier model for resolution.

Agent Dispatcher

Name: Agent Dispatcher
Author: hjemmesidekongen

byhjemmesidekongen

0•

生产力与工作流

Coordinates parallel execution of AI subtasks by assigning worker agents to specific files and model tiers based on project requirements.

The Agent Dispatcher is a sophisticated orchestration layer for Claude Code that manages the transition from task planning to execution. By analyzing decomposed subtasks, it intelligently assigns specialized agents to specific file sets, ensuring no two agents conflict while optimizing performance through tier-appropriate model assignment. It automates the entire worker lifecycle—including parallel execution waves, knowledge loading via reference paths, and a robust retry/escalation logic—to guarantee that complex, multi-part development tasks are completed efficiently and reliably without manual intervention.

主要功能

01Automated retry logic and tier escalation for failed subtasks

02Comprehensive dispatch logging and status tracking for every subtask

030 GitHub stars

04Parallel agent execution with dependency-aware wave scheduling

05Strict file ownership enforcement to prevent merge conflicts between agents

06Context-aware prompt construction using project-specific reference paths

使用场景

01Executing large-scale refactors where multiple independent modules need simultaneous updates

02Automating complex feature development requiring coordination between frontend, backend, and documentation agents

03Managing high-reliability workflows where failed tasks require automatic escalation to more capable AI models

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add hjemmesidekongen/ai agent-dispatcher

For use in Claude.ai and ChatGPT

主要功能

01Automated retry logic and tier escalation for failed subtasks

02Comprehensive dispatch logging and status tracking for every subtask

030 GitHub stars

04Parallel agent execution with dependency-aware wave scheduling

05Strict file ownership enforcement to prevent merge conflicts between agents

06Context-aware prompt construction using project-specific reference paths

使用场景

01Executing large-scale refactors where multiple independent modules need simultaneous updates

02Automating complex feature development requiring coordination between frontend, backend, and documentation agents

03Managing high-reliability workflows where failed tasks require automatic escalation to more capable AI models