Still-chowline394

Ultra Agents — Development skill for Claude Code

Development community

Autonomous multi-round review + fix loop — no per-round approval.

How to install Ultra Agents

Installs to ~/.claude/skills/still-chowline394-magictex-mcp-ultra-agents/SKILL.md

Terminal

mkdir -p ~/.claude/skills/still-chowline394-magictex-mcp-ultra-agents && curl -fsSL https://raw.githubusercontent.com/Still-chowline394/MagicTeX-mcp/HEAD/commands/ultra-agents.md -o ~/.claude/skills/still-chowline394-magictex-mcp-ultra-agents/SKILL.md

Restart Claude Code, or start a new session, for it to be picked up.

What Ultra Agents does

description: Autonomous multi-round review + fix loop — no per-round approval argument-hint: [skill-name] [depth]

Run an autonomous review-and-fix loop on the current paper. Parse **$ARGUMENTS** as up to two tokens: skill name, then depth (both optional, either order is fine to infer from context).

Skill: the named one, else default to academic-paper-revision.
Depth: the number of review→fix rounds to run back-to-back with no approval from me between rounds. Use the giv

Alternatives in Development

Ralph Loop — Autonomous coding loop — works through tasks, commits, repeats 14k ★
Ralphy — My Ralph Wiggum setup, an autonomous bash script that runs Claude Code, Codex, OpenCode, Cursor agent, Qwen & 2.8k ★
/autopilot — Autonomous hunt loop with deterministic scope safety and configurable checkpoints 1.2k ★