MCP server for reading and writing .resx (resource) files with semantic operations
ResxMcp has 4 tools with basic schemas and brief descriptions, but significant quality gaps. All tools are explicitly registered in Program.cs with name, description, and input schema visible in the ToolDesc declarations. However, descriptions are uniformly terse (13-45 chars), parameter descriptions are absent from the schema (only inferred from parameter names), output schemas are not documented, and error handling provides no recovery guidance. No tool annotations (readOnlyHint/destructiveHint/idempotentHint) are present despite clear write/destructive risks. The tools are domain-appropriate for .resx file manipulation, but the implementation falls well short of production quality baselines.
Read .resx as UTF-8 text
Remove a key from .resx
Set or add a key in .resx
Write UTF-8 text to .resx (atomic replace)
Parameter descriptions missing from JSON Schema objects. Tool definitions show parameter names (file, name, value, backup, comment, content) and infer types from context, but ToolDesc class does not carry description strings for individual parameters, only parameter names appear in schema. LLMs cannot determine parameter meaning or constraints from names alone.
Output schemas not documented. No tool response structure is defined. Clients/LLMs cannot determine what fields are returned or how to chain results into downstream calls. Write and setEntry appear to return {ok, file, [name]}, but this is inferred from code, not declared.
Tool descriptions are 22-45 characters, well below the 50-200 char baseline for actionable LLM guidance. 'Remove a key from .resx' and 'Set or add a key in .resx' lack context on when to use each tool, prerequisites, or consequences. Minimal descriptions force LLMs to guess intent.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 44 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 0 | 2024-11-05+ | v1 |
No tool annotations (readOnlyHint, destructiveHint, idempotentHint) present. resx.write and resx.removeEntry have WRITE and DESTRUCTIVE risk but are unmarked. LLMs cannot distinguish safe reads from destructive writes without annotation.
Error handling provides no recovery guidance. Code catches exceptions and returns raw error messages (e.g., ex.Message from file I/O), but doesn't categorize errors as retryable/user-fixable/fatal or suggest next steps. E.g., FileNotFoundException returns '…' without suggesting to verify file path.
Semantic ambiguity between write and setEntry. Both mutate .resx files, but write() replaces entire content while setEntry() upserts a single key. Tool descriptions don't clarify this distinction or guide LLM selection. Without clear separation, agents may misuse write() when setEntry() is safer.
Parameter naming inconsistency: setEntry description says 'key', but schema parameter is 'name'. removeEntry uses 'name'. This forces LLMs to map between conceptually identical fields (key vs. name), increasing confusion and error likelihood.
No input validation guidance. Parameters like 'file' and 'name' have no documented constraints (length, format, allowed characters, path traversal guards). Descriptions don't state whether paths are absolute/relative, whether ~/user expansion is supported, or whether names can contain special chars.