MCP server exposing bodega's yum CLI as native tools, resources, and prompts for Claude Code.
Bodega exposes 12 yum package management tools via TypeScript/MCP SDK. Critical gaps: NO visible input schemas in provided source code (mcp-server/src/tools/*.ts files not shown), descriptions are minimal (10-30 chars, well below 50-200 char baseline), and parameter documentation is absent. Tool names follow verb_noun pattern (yum_install, yum_remove) but lack LLM-optimized descriptions explaining WHEN to use each tool vs alternatives. No evidence of output schema documentation, error recovery guidance, or parameter constraints. The server.ts shows tool registration but actual schema definitions are not visible in the provided code sample.
Find duplicate package installations
View and manage yum operation history
Get detailed information about a package
Install a package via yum
List installed packages or package categories
View the yum operation history log
List packages with available updates
Remove an installed package
Search for packages by name or description
No input schemas visible in source code. Cannot verify parameter types, constraints, or descriptions.
Tool descriptions are 10-30 characters (e.g., 'Install a package via yum'). Baseline is 50-200 chars. Descriptions lack WHEN to use, prerequisites, and differentiation from similar tools (e.g., yum_search vs yum_info).
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 30 | <=2025-11-25 | v2 |
Manage system services installed via yum
Upgrade one or more packages to their latest versions
Verify the integrity of installed packages
No parameter documentation visible. LLMs cannot infer what 'package', 'status', 'filter', or other params mean.
No output schema documentation. LLMs cannot plan downstream tool calls or extract required fields (e.g., does yum_search return package_id, package_name, version?). Breaks tool chaining.
Destructive tools (yum_remove, yum_upgrade) lack error handling guidance and confirmation patterns. No evidence of dry-run, recovery instructions, or actionable error messages.