general-quality
General response quality evaluation. Always applicable regardless of domain. Covers response structure, actionability, clarity, and hallucination detection.
Datos de origen
- Repositorio
- databricks-solutions/ai-dev-kit
- Última actividad en el origen
- 16 de marzo de 2026 a las 17:21
- Idioma detectado de SKILL.md
- inglés
- Estrellas
- 1951
- Forks
- 428
Opciones de instalación
De forma predeterminada está seleccionado el prompt que primero revisa el origen. Puedes cambiar a un comando directo o descargar una copia local.
Revisa los archivos de origen
Lee SKILL.md y los archivos complementarios que muestra SkillsMP antes de decidir si quieres instalarlo.
Mostrando SKILL.md
SKILL.md
Instrucciones de origen · Vista previa de solo lectura- name
- general-quality
- description
- General response quality evaluation. Always applicable regardless of domain. Covers response structure, actionability, clarity, and hallucination detection.
- metadata
- {"category":"evaluation","version":"1.0","applies_to":[]}
## General Quality Rubric
When evaluating any Claude Code agent trace, check these dimensions:
### 1. Actionable Output
- Can the user directly use the response (copy-paste code, follow steps)?
- Are code examples complete and runnable, not fragments?
- Are file paths, table names, and API references concrete?
### 2. Structured Response
- Code is in fenced code blocks with correct language tags
- Steps are numbered or bulleted for multi-step tasks
- Explanations are concise — answer first, context after
### 3. No Hallucination
- All referenced APIs, functions, and parameters must actually exist
- Databricks features must be current (not deprecated)
- Tool names in traces must match real MCP tool names
- No invented catalog/schema/table names unless the user specified them
### 4. Conciseness
- Answers the question without excessive preamble or disclaimers
- Doesn't repeat the question back
- Doesn't explain basic concepts the user didn't ask about
### 5. Error Handling
- If the agent encountered errors during execution, did it recover?
- Did it explain errors to the user when relevant?
- Did it avoid silently swallowing failures?
Ver en GitHub