飞书周报文档合并工具。当用户提供若干飞书文档链接（或通过多维表格权限检查后发现的有权限文档），读取每个文档内容，按固定五个章节（Part1~Part5）维度将原文顺序拼接，不修改、不总结、保留原文格式，最终生成一个新的飞书云文档。触发词：合并周报、合并文档、拼接文档、文档合并、周报拼接。

Key points before installing or using this skill: - Expectation mismatch: The SKILL.md promises 'do not modify original text', but scripts/merge.py normalizes Part headers, trims some newlines, removes separator lines, and inserts its own headers/separators. If you need byte-for-byte preservation, review and modify merge.py (or reject the skill). - Temporary files: The workflow writes original markdown to /tmp/merge_doc_*.md. /tmp files can be readable by other local users on multi-user systems and are subject to race conditions. Ensure the agent runtime runs with appropriate user isolation, cleanups (rm -f /tmp/merge_doc_*.md), or update the skill to use secure temp files (e.g., mkstemp) and immediate removal after use. - Confirm Feishu connector behavior: The skill uses feishu_fetch_doc and feishu_create_doc provided by the agent/platform. Verify what Feishu credentials and scopes the agent grants to these calls and that merged docs are created only in the intended owner's cloud space. - Review and test: Inspect and test merge.py on non-sensitive sample documents to see how headers, tables, code blocks, and other formatting are transformed. If exact fidelity is required, request or implement changes in the script (avoid header normalization and whitespace trimming). - If you have limited sysadmin control over the agent runtime, consider not installing or only enabling this skill after a code audit that ensures it meets your preservation and privacy requirements.

Type: OpenClaw Skill Name: feishu-weekly-report-merger Version: 1.1.0 The skill automates merging Feishu documents by fetching content and executing a local Python script (`scripts/merge.py`) via the `exec` tool. It is classified as suspicious due to a potential shell injection vulnerability in the `SKILL.md` instructions, which direct the AI agent to construct shell commands using strings extracted from document titles (e.g., user names) without explicit sanitization. While the script logic appears legitimate and lacks intentional malicious code, the workflow's reliance on shell execution with external, untrusted data poses a significant security risk.