Loading ToolBench
Code::Stats public coding-XP profiles (codestats.net)
Two tools with minimal definitions. Both have valid input schemas (username: string, required) and basic descriptions, but lack parameter descriptions, output schema documentation, and error recovery guidance. Naming is generic ('user', 'recent') without action verbs. No tool annotations, no pagination support, no field-level documentation. Descriptions are 10 - 70 chars (below the 194-char baseline for A+ tools). Error handling is present but not actionable for LLM recovery.
Recent coding activity (~last 12 hours) for a user.
Public profile — total XP, per-language XP, per-machine XP.
Tool names lack action verbs. 'user' and 'recent' do not convey what the tool does. Should be 'get_user_profile' and 'get_recent_activity' to match LLM expectations.
Parameter 'username' has no description. LLMs cannot infer whether to pass a username, email, or ID. Add: 'The Code::Stats username (case-sensitive, alphanumeric).'
No output schema documented. LLMs do not know what fields to expect (e.g., total_xp, languages, machines). Document the response structure for both tools.
Error messages are generic ('Code::Stats: user not found', 'Code::Stats: 404 ...'). Should guide LLM recovery: 'User not found. Verify the username is correct (case-sensitive) or try search_users() if available.'
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 43 | 2026-07-28+ | v2 |
No tool annotations (readOnlyHint, idempotentHint). Both tools are read-only and idempotent, declare this explicitly for agent planning.