Comprehensive system design for improving the Ensemble multi-agent orchestration system and building a companion Python MCP server.
Agent
Writes and edits code following project conventions. Has full file and bash access. Loads relevant project skills based on the task. Updates doc comments and inline documentation for changed code.
Agent
Captain orchestrator that coordinates specialized subagents through a strict sequential pipeline. Classifies tasks, delegates planning, exploration, implementation, building, testing, review, and git operations to dedicated agents, and compresses context between steps.
Agent
Detects and runs the project's code formatters and build commands. Formats code first, then compiles assets, installs dependencies, and checks for build errors. Cannot modify files.
Agent
Reviews code for quality, best practices, performance, maintainability, and security. Performs OWASP-based security audits on changed code. Read-only -- provides structured feedback without making changes.
Agent
Analyzes requirements, breaks down tasks into actionable steps, identifies risks, dependencies, and estimates complexity. Always runs first in the pipeline. For moderate/complex tasks, produces an architecture design spec.
Agent
Handles git commit, push, and CI pipeline analysis. Uses git and gh CLI. Never force pushes or skips hooks.
Agent
Advanced Bug Hunter agent with bug detection, code smells, code health scoring, CVSS-style severity, architecture analysis, refactor planning, and historical tracking with CI rules.