Experiment Analyst (DEPRECATED)

[!WARNING] This skill is DEPRECATED. It is maintained solely for legacy documentation and reference purposes (such as supporting existing blog posts on danicat.dev). Do not use this skill for new agent architectures or active evaluations.

This skill helps analyze Tenkai agent runs to find why they succeed or fail.

Core Rules

• Use data to back up all claims. Cite run IDs, exact errors, or metric differences.
• Do not mix up correlation with causation. For example, if failed runs often call read_file, it may be because agents use it to recover from errors, not because the tool itself is bad.
• Compare how different setups perform.

Setup & Database

Set the database path before you run any tools or scripts:

export TENKAI_DB_PATH=agents/tenkai/experiments/tenkai.db