Experiment Runner (proj-012) Adaptive experiment runner for survival arena. Runs experiments one by one, analyzes results, commits. Between experiments -- human/AI decides what to change next. Workflow (adaptive) Rule: we only change conditions (pressure, resources, architecture). Never hardcode agent behavior. When to use - "run experiment" / "next experiment" - "what is the experiment status" - "run EXP-011c" - "run the next 2 experiments" - proj-012 experiment pipeline Dependencies - Python 3, PyYAML ( ) - Claude CLI (for LLM queries in the arena) - Git (for committing results) Paths | Wha…