Coding · Intermediate
Benchmark agent skills and tool prompts
Compare agent runs with and without specific tools or skills to measure accuracy
- When something happensTest suite submitted
- ThenRun comparative evals
- ResultGenerate performance diff
Set up automated test suites to measure whether newly added agent skills, instructions, or tools actually improve task completion. Run baseline and skill-augmented runs side by side against test cases, evaluate output quality, and review token usage before deploying changes.
Prompt to start with
BotBentoRecommended
When triggered with an evaluation suite, execute task runs across baseline and skill-enabled configurations using the terminal. Record outputs, score results against expected criteria, and summarize differences in token consumption and error rates into docs. Always wait for user approval before modifying repository configurations or merging prompt updates.