skills.rest
itsadijmbt
itsadijmbtCOMMUNITY
Explore all skills by this author →
GitHub

tool-selection

Evaluate Databricks agent traces for MCP tool selection over Bash workarounds.

1|Updated Apr 24, 2026
Search Tags:#automation#mcp#tool selection#databricks#sql execution#trace evaluation#bash workarounds
One-click install
npx skills add https://github.com/itsadijmbt/SecureMCP-Servers --skill tool-selection-itsadijmbt
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: tool-selection
Source: https://github.com/itsadijmbt/SecureMCP-Servers/tree/main/TEST_SERVERS/PORTED_TO_SECUREMCP/databrickslab-mcp/ai-dev-kit/.test/eval-criteria/tool-selection
Command: npx skills add https://github.com/itsadijmbt/SecureMCP-Servers --skill tool-selection-itsadijmbt

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

This Skill evaluates whether an agent selects the correct Databricks MCP tools instead of relying on Bash shell workarounds, improving tool efficiency and correctness in execution traces.

Core Features & Use Cases

  • MCP Over Bash Enforcement: Flags patterns where equivalent Databricks MCP tools exist but Bash-based alternatives were used instead.
  • Correct-Tool Validation: Ensures operations map to the intended MCP tools (e.g., SQL execution vs notebook execution, workspace reads vs curl).
  • No-Shell-Waivers & Call Efficiency Checks: Detects unnecessary retries and excessive or redundant tool calls that indicate inefficient or confused behavior.
  • Error Recovery Expectations: Verifies that failures trigger alternative approaches rather than blindly repeating the same failing call.

Quick Start

Load this evaluation when you want to score a Databricks agent trace for tool correctness, efficiency, and absence of Bash workarounds.

Frequently Asked Questions about tool-selection

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I evaluate Databricks agent traces for correct MCP tool selection?▼

To enforce MCP over Bash in Databricks, you analyze agent traces to detect when Bash shell workarounds bypass available MCP tools. This Skill flags those patterns and enforces strict mappings for SQL execution, workspace access, and volume file operations.

What is the best way to detect Bash workarounds bypassing Databricks MCP tools?▼

Detecting Bash workarounds bypassing Databricks MCP tools requires scanning agent execution traces for shell commands used instead of native MCP calls. This Skill identifies patterns where curl or Bash scripts replace dedicated MCP operations for SQL, compute, and jobs.

How do I check if my Databricks agent handles tool call failures efficiently?▼

Checking Databricks agent failure recovery involves scanning traces for excessive retries or redundant tool calls. This Skill verifies that failed SQL execution or workspace access attempts trigger alternative approaches rather than blindly repeating the same call.

Can I use trace evaluation to validate SQL execution tool choice in Databricks agents?▼

Trace evaluation validates SQL execution tool choice in Databricks agents by mapping logged operations to intended MCP tools. This Skill ensures SQL queries and notebook executions route through correct MCP interfaces rather than Bash shell alternatives.

Why does my Databricks agent use Bash instead of MCP tools for workspace access?▼

Databricks agents may use Bash instead of MCP tools for workspace access when execution traces show tool selection drift toward shell workarounds. This Skill identifies those instances and enforces correct MCP mappings for workspace reads, volume file operations, and compute management.

Related Skills

Explore More Data & Analytics Skills→
leary-pokenleary-poken
1

tool-selection

Evaluate agent traces to flag Bash workarounds and recommend Databricks MCP alternatives.

Community
Intermediate
jaydubya818jaydubya818

evaluation

Define rubrics, manage test sets, and analyze agent performance.

Community
Advanced
vuminhieuvuminhieu

mcp-management

Discover, analyze, and execute MCP capabilities across configured servers.

Community
Advanced
ngquaanghuyngquaanghuy

skill-security-auditor

Analyze AI agent skill files for command injection, data exfiltration, and prompt injection.

Community
Advanced