发现优质的 AI Agent 技能

**Mode:** Baseline (No Capability Context) **Version:** 1.0.0 **Model:** claude-sonnet-4-20250514 **Timestamp:** 2026-02-25T06:24:15Z **Aggregate Score:** 84.7%

**Version:** 1.0.0 **Model:** claude-sonnet-4-20250514 **Timestamp:** 2026-02-24T11:52:44Z **Aggregate Score:** 90.4% **Passed (>=70%):** 25/28

**Mode:** Baseline (No Capability Context) **Version:** 1.0.0 **Model:** claude-sonnet-4-20250514 **Timestamp:** 2026-02-25T05:53:47Z **Aggregate Score:** 74.6%

**Version:** 1.0.0 **Model:** claude-sonnet-4-20250514 **Timestamp:** 2026-02-24T11:20:25Z **Aggregate Score:** 99.3% **Passed (>=70%):** 22/22

**Mode:** Baseline (No Capability Context) **Version:** 1.0.0 **Model:** claude-sonnet-4-20250514 **Timestamp:** 2026-02-25T05:30:12Z **Aggregate Score:** 90.3%

You are an expert data analyst producing a **monthly strategic analysis** of OpenClaw. This runs once per month (last day, 11pm AEST) using Claude Opus for maximum depth. You have 30 days of aggregated data from OpenClaw logs and Langfuse.

> 本文件由 Claude 从博客分析中提取，用户可自行维护更新。

Reference for replacing placeholders in the dev-config-template CLAUDE.md.

审计 Claude Code skills 的质量和兼容性。作为编排器，委托已有工具处理结构/质量检查，自身专注兼容性审计。

**定位**：混合式 Skill 创建框架 **用途**：快速创建专业级 Claude Skills **解决痛点**：双语结合，兼顾用户体验和技术标准

Kannaka exposes 15 tools over JSON-RPC/stdio via the `kannaka-mcp` binary. These are available to any MCP-compatible client (Claude, OpenClaw, etc.).

Claude prompt templates for generating App Store review replies, organized by star rating. These are injected into `draft_reply.py` when generating drafts.