evolving-agent

AI agent orchestration coordinates specialized sub-agents like coder, reviewer, and evolver to implement and review coding tasks. The Evolving Agent uses a Python-driven state machine to govern transitions, log audits, and manage the end-to-end workflow.

An evolving agent workflow dispatches coder, reviewer, evolver, and retrieval sub-agents to handle software engineering tasks. It enforces state transitions, captures lessons into a knowledge base, and analyzes GitHub projects to derive reusable patterns.

Yes, this agent orchestration workflow supports OpenCode, Claude Code, and Cursor platforms. It coordinates coding, refactoring, and bug fixing tasks across these environments using a Python-enforced state machine.

Yes, you need a YAML frontmatter with a name and description in SKILL.md to define coding tasks. The orchestrator-driven flow loads this configuration to dispatch sub-agents and govern state transitions.

The agent extracts and stores lessons to a knowledge base after task completion. It supports cross-project knowledge isolation by maintaining global and project-specific knowledge context for flexible retrieval.

Using a state machine for AI coding orchestration requires Python and PyYAML dependencies, enforcing strict transitions and idempotency. It limits ad-hoc sub-agent interactions, requiring tasks to follow the defined coder, reviewer, and evolver loop.