Thinking - 深度思考与规划

通过系统化元认知流程，深入解析任务、盘点技能、检索经验、规划路径与评估风险，提升复杂任务执行效果与效率。

This skill is an instruction-only meta-skill that will ask the agent to read its own memories, session history, and available skill list to form plans — that behavior is expected for a planning tool but can surface private information from your agent's memory and conversation history. Before installing: 1) confirm you are comfortable with other installed skills and stored memories being consulted; 2) remember the skill itself does not send data externally, but examples (e.g., posting to Feishu) would require calling other skills/services that may need credentials — review and limit those integrations; 3) note the package.json references a src/index.ts that isn't included, which is acceptable for an instruction-only skill but worth being aware of. If you want stricter limits, restrict the agent's permission to access memory/session history or avoid granting downstream skills (e.g., messaging integrations) access to sensitive destinations.

Type: OpenClaw Skill Name: thinking-skill Version: 1.1.0 The 'thinking-skill' bundle is a metacognitive framework designed to guide an AI agent through a structured six-step reasoning process (Task Analysis, Skill Inventory, Memory Retrieval, Path Planning, Risk Assessment, and Execution). The instructions in SKILL.md and README.md are entirely consistent with the stated goal of improving agent planning and reliability. There is no evidence of malicious intent, data exfiltration, or harmful prompt injection; in fact, the skill explicitly includes a 'Risk Assessment' step to identify potential API or permission issues before execution.