What are the minimum test requirements for a reviewer agent?

Reviewer agents require at least 6 test cases: 2 known issues, 2 clean inputs, 1 edge case, and 1 ambiguous case to ensure full coverage.

What is the difference between testing skills and testing agents?

Skills are instruction sets that guide behavior, whereas agents are separate Claude instances performing specific tasks that require rigorous output verification.

Why is the Iron Law of testing so strict?

Untested agents often fail in production due to unexpected edge cases or hallucination; a 20-minute test suite prevents hours of debugging and lost trust.

Can I skip testing for simple agents?

Only with explicit human partner approval. Even simple agents can behave unpredictably, so the default requirement is always to test.

Agent Testing with Subagents

Name: Agent Testing with Subagents
Author: LerianStudio

byLerianStudio

•

安全与测试

Implements a strict Test-Driven Development (TDD) methodology for AI agents to ensure accurate, well-structured, and reliable outputs.

关于

The Testing Agents with Subagents skill applies the RED-GREEN-REFACTOR cycle specifically to the development and deployment of AI workers. It enforces a mandatory testing workflow where agents must be validated against known inputs, edge cases, and accuracy metrics before deployment. This skill is critical for developers building complex multi-agent systems where subagents perform specialized tasks like code review, data analysis, or decision-making, ensuring they adhere to schemas and maintain consistency across parallel executions.

主要功能

Standardized RED-GREEN-REFACTOR workflow for prompt engineering
Detailed accuracy metrics including True/False Positive rates
17 GitHub stars
Edge case and stress testing protocols for robustness
Consistency verification for parallel agent workflows
Comprehensive test suite requirements for diverse agent types

使用场景

Validating code reviewer agents against known security vulnerabilities
Calibrating severity levels for automated triage and decision agents
Testing structured data extractors for strict schema compliance

关于

主要功能

Standardized RED-GREEN-REFACTOR workflow for prompt engineering
Detailed accuracy metrics including True/False Positive rates
17 GitHub stars
Edge case and stress testing protocols for robustness
Consistency verification for parallel agent workflows
Comprehensive test suite requirements for diverse agent types

使用场景

Validating code reviewer agents against known security vulnerabilities
Calibrating severity levels for automated triage and decision agents
Testing structured data extractors for strict schema compliance