Skip to content

fix: update skills to use execute_code MCP tool name#385

Merged
calreynolds merged 1 commit intomainfrom
fix-compute-tool-name
Mar 27, 2026
Merged

fix: update skills to use execute_code MCP tool name#385
calreynolds merged 1 commit intomainfrom
fix-compute-tool-name

Conversation

@QuentinAmbard
Copy link
Copy Markdown
Collaborator

Summary

  • Skills were incorrectly referencing internal function names (execute_databricks_command, run_file_on_databricks) instead of the actual MCP tool name execute_code
  • This caused LLMs to attempt using non-existent tools
  • Updated all skill documentation to use correct MCP tool names
  • Applied ruff auto-fixes across the codebase

Changes

Skills updated:

  • databricks-model-serving - all references now use execute_code
  • databricks-zerobus-ingest - all references now use execute_code

Other updates:

  • system_prompt.py - correct tool references
  • README files - updated tool documentation
  • Test manifests - updated tool limits

MCP Tool Mapping

The MCP server exposes 4 consolidated compute tools:

MCP Tool Purpose
execute_code Unified entry point for code/file execution
manage_cluster Cluster lifecycle management
manage_sql_warehouse SQL warehouse management
list_compute List clusters, node types, spark versions

Test plan

  • Tests pass (41 passed)
  • Imports compile successfully
  • Ruff linting applied

🤖 Generated with Claude Code

@QuentinAmbard QuentinAmbard force-pushed the fix-compute-tool-name branch from 077d9a2 to c6a224e Compare March 27, 2026 09:41
Skills were incorrectly referencing internal function names like
`execute_databricks_command` and `run_file_on_databricks` instead of
the actual MCP tool name `execute_code`. This caused LLMs to attempt
using non-existent tools.

Also removes the backwards-compatibility alias `run_python_file_on_databricks`
from databricks-tools-core since it's not exposed via MCP.

Changes:
- Updated databricks-model-serving skill to use `execute_code`
- Updated databricks-zerobus-ingest skill to use `execute_code`
- Updated system_prompt.py, README files, and DocPage.tsx
- Removed `run_python_file_on_databricks` alias from compute module

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>
@QuentinAmbard QuentinAmbard force-pushed the fix-compute-tool-name branch from c6a224e to 66cb710 Compare March 27, 2026 09:50
@calreynolds calreynolds merged commit 2aa1582 into main Mar 27, 2026
SarojDB pushed a commit to SarojDB/ai-dev-kit that referenced this pull request Apr 2, 2026
…tions#385)

Skills were incorrectly referencing internal function names like
`execute_databricks_command` and `run_file_on_databricks` instead of
the actual MCP tool name `execute_code`. This caused LLMs to attempt
using non-existent tools.

Also removes the backwards-compatibility alias `run_python_file_on_databricks`
from databricks-tools-core since it's not exposed via MCP.

Changes:
- Updated databricks-model-serving skill to use `execute_code`
- Updated databricks-zerobus-ingest skill to use `execute_code`
- Updated system_prompt.py, README files, and DocPage.tsx
- Removed `run_python_file_on_databricks` alias from compute module

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-authored-by: Quentin Ambard <quentin.ambard@databricks.com>
Co-authored-by: Claude <noreply@anthropic.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants