poly-mcp has 6 tools with basic schemas and descriptions, but significant quality gaps. Tool names lack clear action verbs (workspace_lint vs lint_workspace), descriptions are minimal (10-50 chars, well below the 194-char baseline), and parameter descriptions are sparse or missing. No input validation guidance, no error recovery hints, and no output schema documentation. The server exposes file paths as parameters without sanitization guidance, raising injection risks. Tool composition is reasonable (separate lint/format/config tools), but descriptions don't explain when to use each or what they return.

• Expand tool descriptions to 100-200 chars. Example: 'Lint specified files and optionally apply fixes. Returns a list of issues (file, line, rule, severity) and fix counts. Use lint_files for targeted checks; lint_workspace for full project analysis.'
• Add parameter descriptions: 'paths: Array of file paths or glob patterns (e.g. ["src/**/*.ts", "tests/"]). Relative to workspace root. Required.' and 'fix: If true, apply auto-fixes to linting issues. Defaults to false (dry-run).'
• Document output schemas for each tool. Example: 'Returns {issues: [{file, line, column, rule, severity, message}], fixed_count: number, error?: string}'
• Rename workspace_lint and workspace_lint_fix to lint_workspace and lint_workspace_fix for consistent verb_noun naming.
• Add error recovery guidance: 'If a file is not found, returns {error: "File not found: <path>"}. Check paths are relative to workspace root.' This helps LLMs self-correct.