Workspace Guardian

文件创建与目录组织规则。强制执行统一的命名规范和目录结构，确保所有产出物可预测、可查找、可追溯。价值：节省 token（减少重复搜索定位）、节约时间（不再花时间找文件）、降低管理成本（产出物有序不混乱）。当用户没有明确指定路径时，此规则强制生效。

This skill is coherent with its stated goal of enforcing file naming and placement, but it contains rules that automatically delete or move files (including deleting failed scripts and clearing temp/empty directories) and states the rule 'overrides default behavior' when a user does not provide a full path. Before installing or enabling it, consider these precautions: - Require explicit confirmation for any destructive action (deletion or permanent archive) rather than immediately acting on short user phrases like "不好看" or "再生成一个". - Limit the agent's write/delete scope to quarantined/specific workspace directories (e.g., only under ~/WorkBuddy/* or a user-approved folder) and disallow sweeping operations across the entire home directory. - Add a dry-run/log-only mode so the agent reports proposed deletions/moves and waits for approval. - Ensure actions are logged and that there is an easy undo/recovery path (e.g., move to a quarantine folder instead of immediate permanent deletion for N days). - If you accept automated cleanup (temp/24hr rules), back up important files first and test the skill in a sandbox environment. - Ask the author (or require modification) to make destructive rules opt-in and to clarify ambiguous triggers for deletion (what precisely counts as '部分成功' vs '致命错误'). If you cannot enforce these safeguards, treat the skill as potentially risky and avoid granting it autonomous use over your primary home/work directories.

Type: OpenClaw Skill Name: workspace-guardian-by-lujun Version: 1.0.2 The skill bundle contains instructions in SKILL.md that implement an aggressive 'Automatic Cleanup Mechanism,' directing the AI agent to autonomously delete files, scripts, and directories based on subjective user feedback (e.g., 'not good' or 'redo') or script failures. While the stated goal is workspace organization, the use of authoritative language to override default agent behavior and the potential for unintended data loss via automated file deletion based on natural language interpretation are high-risk behaviors.