Loading ToolBench
Code snippet manager with language tags and search for AI agents. Reads free, writes require Veyra.
Server has 6 tools with complete schemas and descriptions, but quality is uneven. All tools have input schemas with type definitions and descriptions present. However, descriptions are inconsistent in depth, some are generic ('List code snippets') while others include pricing info ('Requires Veyra commit mode'). Parameter descriptions are minimal (5-15 chars typical). No output schemas documented. Error handling is basic (JSON error objects without recovery guidance). Missing tool annotations (readOnlyHint, destructiveHint). Naming is clear and verb-based (list_, get_, search_, save_, update_, delete_). No pagination support despite list_snippets potentially returning many results. Veyra token handling exposes a security concern: tokens passed as parameters rather than server-side injection.
Delete a snippet by ID. Requires Veyra commit mode (Class B — €0.02).
Retrieve a code snippet by ID. FREE — no token required.
List code snippets, optionally filtered by language or tag. FREE — no token required.
Save a new code snippet. Requires Veyra commit mode (Class A — €0.005).
Search snippets by query (matches title, code, language, tags). FREE — no token required.
Update a snippet's code or title. Requires Veyra commit mode (Class A — €0.005).
Veyra tokens exposed as tool parameters instead of server-side secret injection. Tokens will appear in agent logs and traces, creating a security leak.
No output schemas documented. LLMs cannot predict response structure (fields, types, nesting). Responses are JSON strings, not structured objects.
Missing tool annotations. No readOnlyHint on read tools (list_, get_, search_) or destructiveHint on delete_snippet. Agents cannot distinguish safe from risky operations.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 56 | 2026-07-28+ | v2 |
list_snippets lacks pagination (limit, offset, page_size). No total count or next_cursor returned. Large result sets will blow context windows.
Error responses are bare JSON objects without recovery guidance. 'SnippetNotFound' tells agent nothing about next steps. Should suggest search_snippets or list_snippets.