Multi-agent
What is Testing Agent?
A testing agent exercises another system and reports pass, fail, or regressions.
It is an evaluator with executable checks, not a reviewer of prose alone.
Why it matters
A testing agent runs executable checks against another system and reports pass, fail, or regressions. Unlike evaluators that assess quality, testing agents produce binary or graded results backed by execution.
Testing agents close the verification loop — they provide evidence, not opinion.
Key takeaways
- 1Testing agents produce executable, evidence-backed results.
- 2They provide verification, not opinion.
- 3Tests should cover expected behavior and known failure modes.
Common mistakes
- ✕Running tests without assertions — execution without expected outcomes is not testing.
- ✕Not including negative test cases (expected failures).
Related terms
Concept neighborhood
Terms linked from Testing Agent in the glossary graph.
- Testing Agent
- Agent Evaluation
- Verification
- Eval Harness