zenith

name	zenith
description	Runs a long-running mission using the Zenith continuous-improvement harness to dynamically orchestrate workers, testing, and stopping discipline.

Zenith: Long-Running Task Harness

When this skill is invoked:

1. First read .agents/orchestrator_prompt.md and treat it as your primary role.
2. Use Zenith to run this mission.

About the Zenith Harness

Zenith is a continuous-improvement harness designed specifically for long-horizon, complex tasks where the primary failure mode is premature completion. It uses an adaptive control loop to dynamically allocate workers, run validators, update task lists, and make authoritative stopping decisions.

Setting Up Zenith in a Workspace

Initialize the workspace using the Zenith CLI:

# Run from the local share zenith checkout:
uv run zenith init --workspace-dir /Users/albertonunez/Documents/Developer/BurnBar --agent antigravity

This writes .mcp.json with the Zenith stdio server registration, and stages the orchestrator system prompt at .agents/orchestrator_prompt.md.

Running Zenith

Start your agent from the workspace:

# Verify the zenith MCP server is running in the agent environment.
# Then ask the agent:
First read .agents/orchestrator_prompt.md and treat it as your primary role, then use Zenith to run this mission.
<your instructions/query>