Loading ToolBench
Control Claude Code from your phone via Discord. Bridges Claude Code sessions to Discord, allowing remote interaction through Discord messages, file uploads, and interactive buttons.
Two tools with minimal descriptions (both under 50 chars) and incomplete parameter documentation. send_to_discord lacks description of what 'caption' does or file format constraints. react_in_discord has no guidance on valid emoji formats. Neither tool documents output structure, error cases, or recovery paths. Parameter descriptions are present but generic. No input validation rules, no error handling guidance, no idempotency guarantees documented.
React with emoji in the Discord session
Send files or images to the Discord session
Tool descriptions are too short (both <50 chars). LLMs cannot determine when to select these tools vs alternatives or what side effects occur. Descriptions must explain WHAT, WHEN, and consequences.
No output schema documented. LLMs cannot plan downstream actions or extract success/failure signals. Both tools must return structured responses (e.g., {success: bool, message_id?: string, error?: string}).
No error handling guidance. If send_to_discord fails (invalid file, Discord API error), the LLM has no recovery path. Errors must be actionable: 'File not found at /path/to/file. Check the path and retry.' not bare exceptions.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 46 | 2026-07-28+ | v2 |
Parameter 'emoji' in react_in_discord lacks format constraints. Valid emoji formats (Unicode, Discord emoji ID, emoji name) are undocumented. LLMs will guess and pass invalid values.
send_to_discord 'caption' parameter description is generic ('Optional caption for the file'). Does it support markdown? Max length? Required for images but not files? Ambiguity forces LLM guessing.