/go — Goal-Driven Entry Point

What it does

This skill implements a single /go entry point for an agent conductor/orchestrator system. Users state a goal and the skill parses intent, detects or creates a track, generates spec and plan documents, emits track metadata, and dispatches an evaluate-loop that coordinates multiple specialized agents (planning, execution, debugging). It automates track lifecycle: create track directories, produce spec.md and plan.md (DAG), generate metadata.json, and run monitoring and quality gates.

When to use it

Use /go when you want to convert a high-level request into a managed multi-step project: feature development, bugfixes, refactors, or larger tasks that benefit from an agentic evaluate-loop. It is appropriate when you want autonomous orchestration or to resume existing tracks. Not intended for trivial single-step commands.

What's included

• Scripts: none detected in skill folder (expects conductor orchestration components elsewhere in the repo)
• References: none embedded in the skill file; relies on repo-level tracks.md and config files (e.g., conductor/config.json)
• Instructions: procedural guidance for goal analysis, track detection/creation, metadata schema, and dispatch rules for superpower-enhanced vs legacy loops.

Compatible agents

Designed for multi-agent orchestration platforms (Claude Code, Claude-based orchestrators) and agents that support running evaluate-loop style workflows and interacting with repo file systems. Works best when integrated with conductor-orchestrator components.