AgentsAvançado

Testing Skills With Subagents

porneolabhq·neolabhq· v1.0.0 · atualizado em 2026-04-10

Score

Use this skill during skill creation or editing, prior to deployment, to verify its robustness and resistance to rationalization. It applies a RED-GREEN-REFACTOR cycle to process documentation by establishing a baseline without the skill, developing the skill to address failures, and iteratively refining it to close loopholes.

ai-agent-testingtest-driven-developmentred-green-refactorskill-validationpressure-testingrationalization-preventionagent-compliance

Linguagens

TypeScriptJavaScriptPythonJavaC#

0Stars

0Forks

0Usos

Fork

Documento do Skill

SKILL.mdcustomaize-agent-test-skill/workflow

RED Phase (Baseline Testing): — Create pressure scenarios and run them without the skill, documenting agent failures and rationalizations.

GREEN Phase (Skill Writing): — Write the skill to address the specific failures observed in the RED phase.

Verify GREEN (Pressure Testing): — Run the same scenarios with the skill enabled to ensure compliance.

REFACTOR Phase (Loophole Closing): — Identify new rationalizations, add explicit counters, and update the skill.

Re-verify: — Re-test the scenarios to ensure the skill remains compliant after refactoring.

Meta-Testing: — If the agent still fails, ask how the skill could be improved to prevent the violation.

Iterate: — Repeat the REFACTOR and Re-verify steps until the skill is bulletproof.

Telemetria de Agentes

Execuções

total

Taxa de Sucesso

últimos 30d

Latência Média

0.0s

p50

Alucinação

0.0%

detecção

Tokens Entrada

avg 0/exec

Tokens Saída

avg 0/exec

Uso por Plataforma

Skills Relacionados

Árvore do Skill

Testing Skills With Subagents

customaize-agent-test-skill

Fases Cognitivas5

1.SENSE

2.CONTEXTUALIZE

3.EVALUATE

4.ACT

5.REFLECT

Triggers8

test an AI agent skillvalidate an AI skillensure an agent follows rulesprevent agent rationalizationapply TDD to skill developmentverify skill compliance under pressureclose loopholes in a skillrun baseline tests for a skill

Avaliar este Skill

Score Breakdown

⭐Avaliação Humana0%

🤖Sucesso de Agentes0%

🕐Atualidade100%

🔗Saúde de Dependências100%

🕸️Centralidade no Grafo0%

🛡️Segurança50%

CompositeScore = α·Humano + β·Agente + γ·Recência + δ·Deps + ε·Centralidade + ζ·Segurança

Instalação

$ synaptic mcp download customaize-agent-test-skill

$ synaptic skills detail customaize-agent-test-skill

$ synaptic skills live customaize-agent-test-skill

Links

GitHub Repository