train

name

train

description

Autonomous agent and skill trainer — transforms the current Claude session into a self-improving loop inspired by Karpathy's autoresearch. Activates on: "train", "evolve agent", "improve agent", "run training loop", "agent trainer", "evolve this agent", "train this agent", "self-improving agent", "auto-improve", "training loop", "agent evolution", "autotrainer", "improve skill", "evolve skill", "optimize skill", "skill training", "improve my skills". Targets agents (.claude/agents/) OR skills (.claude/skills/) — auto-detects which. For agents: spawns A/B variants, evaluates output quality, mutates config. For skills: uses Claude Code's built-in eval system (run_eval.py) to measure trigger accuracy, then mutates the skill description. Use when you want an agent or skill to improve itself over N cycles without manual intervention.

Train — Autonomous Agent & Skill Trainer

What This Does

The /train skill turns Claude into an autonomous training loop for agents OR skills. It follows the autoresearch pattern: evaluate → reflect → mutate → re-evaluate → keep/revert → repeat.

Agent mode: Spawns A/B subagent variants, evaluates output quality (binary + qualitative), mutates the agent config, keeps winners.

Skill mode: Uses Claude Code's built-in eval system (run_eval.py) to measure trigger accuracy, mutates the skill description, keeps improvements. Can also generate evals if none exist.

Variables

Parameter	Flag	Default	Description
Target	--target <name>	test-subject	Agent name OR skill name to evolve
Mode	--mode <agent|skill>	auto-detect	Force agent or skill mode (auto-detects by default)
Review interval	--review-every <N>	10	Pause for human review every N cycles
Pacing strategy	--pace <fast|medium|deep|adaptive>	adaptive	How deeply to evaluate per cycle
Max cycles	--cycles <N>	0 (infinite)	Stop after N cycles; 0 = run until stopped

Target Resolution

The trainer auto-detects what you're pointing it at:

1. Check .claude/skills/{TARGET}/SKILL.md — if exists → skill mode
2. Check .claude/agents/{TARGET}.md — if exists → agent mode
3. If both exist, use --mode to disambiguate
4. If neither exists, halt with error

Instructions

1. Parse parameters from the args string passed to this skill:

   • Extract --target, --mode, --review-every, --pace, and --cycles values
   • Fall back to defaults for any parameter not provided

2. Resolve target using the Target Resolution rules above.

   • Determine MODE: agent or skill
   • Display: "Target: {TARGET} | Mode: {MODE}"

3. Read LOOP.md from the same directory as this skill file:

   • Path: .claude/skills/train/LOOP.md (relative to project root)

4. Follow LOOP.md instructions using the parsed parameter values.

   • Pass all parameters into the loop as context, including MODE
   • LOOP.md handles both agent and skill training loops