Conjure a Experimentalist Agent

SkillAI & models

Conjure a experimentalist agent — initialize and launch a new autonomous experiment

Available today. Use it from your connected AI after setup.

No other account needed.

Connect ahel once, and every AI you use reads what you have installed.

Then ask your AI: use the Conjure a Experimentalist Agent skill

What this skill tells your AI

The instructions your AI receives, as published by rlacombe/distillate in .claude/skills/conjure/SKILL.md and read by ahel’s review.

Summon a new experiment agent into being. From an empty flask (directory) to a living, running experimentalist agent.

The user provides a research goal, and optionally a directory path, constraints, or a template name.

Gather requirements: Ask the user for:
- Research goal / hypothesis
- Experiment directory path (or create a new one under EXPERIMENTS_ROOT)
- Time budget per iteration (default: 5 minutes) — controls the MAX_SECONDS guard
- Constraints (max sessions, hardware, model preference)
- Whether to base on a template (mcp__distillate__list_templates) or start fresh
Initialize: Call mcp__distillate__init_experiment with the path and goal
- This scans the directory, drafts PROMPT.md with Claude, sets up hooks and tracking
- Review the generated protocol with the user
Set goals: Call mcp__distillate__update_goals to set measurable success criteria
- Pull baselines from literature if relevant (mcp__distillate__extract_baselines)
Launch: On user approval, call mcp__distillate__manage_session with action="start"
- The experimentalist agent comes alive in a tmux session
Verify: Call mcp__distillate__manage_session with action="status" to confirm it's running
Create GitHub repo: Offer to call mcp__distillate__create_github_repo for public tracking

Advanced