agent-test-harness

To test if an AI subagent triggers correctly, you need a test harness that executes representative prompts multiple times to statistically measure trigger accuracy and identify false positives. This skill automates that validation process.

Cross-trigger interference in skill prompt evaluation occurs when multiple skills incorrectly fire on the same input. This skill detects these conflicts by running prompts across multiple trials and reporting per-skill trigger data.

You can measure the statistical significance of skill description tuning by running a test prompts file through this harness. It executes each prompt at least three times to ensure reliable trigger-rate reporting for tuning opportunities.

Yes, you can evaluate live skill frontmatter without modifying definitions by using a read-only test harness. This skill reads current names and descriptions during evaluation and focuses strictly on measurement and reporting.

Skill testing requires multiple runs per prompt for statistical accuracy because single trials can produce inconsistent triggers. This harness executes each test prompt three or more times to ensure statistically significant results.