AIFIRST burial-depth: A Tool That Measures How Buried an Agent's Answer Actually Is
The brief
burial-depth measures the number of words of preamble before an AI agent's response reaches its actual answer, producing a score that can be tracked across model versions and prompt configurations.
Key points
- It is designed to quantify and test verbosity drift in agent systems.
- The tool can be used in CI pipelines to enforce response directness thresholds.
Sources
- GitHubgithub.com
- Source codegithub.com