researcher

Autonomous experimentation works by coordinating planning, execution, and logging within a local lab workspace. It guides iterative think, test, and reflect cycles while enforcing safety guardrails until a measurable target is achieved.

The think test reflect cycle is an iterative experimentation loop that logs results and enforces guardrails for safe testing. It coordinates planning, execution, and reflection to drive continuous learning until a measurable target is reached.

Yes, autonomous experimentation works across any domain that produces measurable results. It uses a lab mindset to orchestrate iterative hypothesis testing, branching, and logging regardless of the specific research field.

A SKILL.md file requires YAML frontmatter containing name and description fields at the root directory. Optional scripts, references, and assets can be included alongside the file while adhering to token hygiene and safety rules.

Guardrails limit uncontrolled execution by enforcing safety rules and token hygiene during autonomous experiments. They restrict subagent collaboration and branching to operations recorded within the local lab history to ensure auditability.