Loading ToolBench
飞书 Channel for Claude Code — forwards Feishu group messages into Claude Code sessions and relays permission requests to Feishu approvals. Reuses local lark-cli authentication without creating separate Feishu apps or handling token refresh.
Single tool 'lark_reply' has a minimal schema with only one parameter. Description is present but generic (27 chars). Parameter 'text' lacks a description entirely, violating the requirement that every parameter must have one. No output schema documented. No error handling guidance. Tool is write-only with no recovery patterns.
把消息发回飞书群（用户正在手机上看）
Parameter 'text' has no description. Rubric requires every parameter to have a non-empty description explaining what it controls.
Tool description is only 27 characters ('把消息发回飞书群（用户正在手机上看）'). Rubric baseline is 194 chars average; descriptions under 20 chars score 0-20. This one is marginally above but lacks WHAT, WHEN, and prerequisites detail.
No output schema documented. Rubric requires documentation of return types so LLMs know what fields to expect. Current code returns {type:'text', text:'sent'} but this is not declared in the tool definition.
No error handling guidance. Tool can fail silently (send() catches errors and logs only). LLM receives 'sent' regardless of actual delivery success, violating recovery-guide pattern.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 44 | 2026-07-28+ | v2 |
Tool is destructive (sends messages to external system) but has no confirmation step or dry-run mode. Rubric pattern:confirmation-request recommends confirmation for irreversible operations.