[Feature] Safe execution skill for MCP actions (pre-validation & guardrails) #372
Description
I’ve been exploring the AI Dev Kit and noticed strong capabilities around execution via MCP tools.
One potential gap is pre-execution validation and guardrails.
I’d like to propose a new skill such as databricks-safe-execution that would:
- validate resource existence before creation
- detect potential duplication or overwrite risks
- simulate actions (dry-run)
- suggest safer alternatives before executing MCP tools
This could help prevent issues like duplicate resource creation and improve reliability when using AI agents in production-like environments.
Happy to draft a first version of the skill if this aligns with the project direction.