Loading ToolBench
Migration and configuration management tools for agent skills across multiple IDEs (Cursor, Claude, VS Code, Windsurf, JetBrains, and others). Provides path resolution, MCP configuration migration, and skill/rule/prompt file management.
This bash-based MCP server has severe definition quality gaps. Tool descriptions exist but are often vague or incomplete (e.g., 'migrate configuration, MCP servers, skills, rules, and prompts' lacks WHEN/WHY context). Input schemas are present but many parameters lack proper type constraints, 'objects' and 'strategy' are free-form strings instead of enums, inviting hallucinated values. No output schemas are documented. Error handling is absent from descriptions. Parameters like 'workspace' lack format guidance. The server conflates multiple concerns (legacy-migration does 6+ things). No tool annotations (readOnlyHint/destructiveHint) despite WRITE operations. Bash implementation makes schema introspection impossible, tool definitions are inferred from script arguments, not explicitly registered, capping per-tool scores at 50.
Apply a migration plan with support for safe mode, lossy content handling, and conflict resolution
Migrate configuration, MCP servers, skills, rules, and prompts between IDEs with support for dry-run, backup, and overwrite strategies
Convert MCP server configurations between IDE formats (mcpServers, mcp.servers, etc.)
Copy VS Code prompt files to target IDE format
Convert and migrate rule/instruction files between IDE formats (MDC, Markdown, AGENTS.md, etc.)
Copy and validate agent skill directories between IDE locations
Print the canonical path for a given IDE and object type (global, project, project-skills, rules, mcp, project-mcp, config, project-config)
Validate IDE path mappings against ide-paths.json reference
No output schemas documented. LLMs cannot infer what fields to expect from tool responses, forcing unstructured parsing and increasing hallucination risk.
Parameters 'objects', 'strategy', 'scope', 'source_format', 'target_format' are free-form strings instead of enums. LLMs will hallucinate invalid values like 'migrate-all' or 'smart-copy' instead of using documented options.
Tool 'legacy-migration' combines 6+ responsibilities (migrate config, MCP, skills, rules, prompts, agents). Should split into separate tools so each has one clear job per pattern:tool.
No error handling guidance in descriptions. When migration fails, LLMs have no recovery path. Descriptions must state: 'If source IDE not found, call validate-ide-paths first.'
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 39 | 2026-07-28+ | v2 |
No tool annotations (readOnlyHint/destructiveHint/idempotentHint). Tools like 'legacy-migration' and 'migrate-skills' perform WRITE operations but lack destructiveHint, preventing agents from understanding irreversibility.