Install
openclaw skills install @harrylabsj/metacognition-checkerCheck whether a user truly understands a concept, decision, plan, or belief, or is only feeling familiar and overconfident. Use for learning validation, decision review, blind-spot detection, error review, confidence calibration, and minimum verification planning.
openclaw skills install @harrylabsj/metacognition-checkerUser input: "I think I understand recursion, but I'm not sure if I really get it" Expected output: The skill runs the four checks (Explain, Transfer, Boundary, Evidence). Identifies whether the user has familiarity vs true understanding. Outputs the minimal verification action: "Close all references and write 3 sentences explaining recursion in your own words, then give one novel example."
User input: "I'm confident this business plan will work. My evidence is market research and competitor analysis. Check my reasoning." Expected output: The skill reveals the current claim, separates facts from interpretation, identifies the most likely blind spot (e.g., survivorship bias in competitor data), and provides one minimum verification action before proceeding.
User input: "I made a wrong judgment call in that meeting. I thought the client would prefer option A but they chose option B. Help me analyze what went wrong." Expected output: The skill separates facts from interpretation, identifies likely bias (e.g., confirmation bias from past experience), and provides a structured breakdown of what was known vs assumed.
帮助用户检查自己的判断、学习或决策过程,识别"感觉懂了"与"真的懂了"的差异。
User input: "我在准备考研,每天学8个小时但感觉什么都没记住,是不是学习方法有问题?" Expected output: 用元认知检查法分析学习效率:1)识别'虚假学习'——被动看视频/划书不等于主动学习;2)推荐的主动学习方法:费曼技巧(用自己的话讲出来)、间隔重复(艾宾浩斯遗忘曲线)、自我测试(合上书回忆);3)提供学习策略诊断清单:是否做笔记?是否会做思维导图?是否会定期复习?建议换用番茄工作法(25分钟学习+5分钟休息),其实真正的有效学习时间可能只有4小时。
根据用户给出的判断、计划或理解,输出:
Use these checks before giving advice:
If one check fails, recommend a minimum verification action instead of continuing with abstract reassurance.
Keep the response short unless the user asks for a full audit:
<判断><事实><解释><盲点/反例><最小验证动作>用户:我觉得我已经学懂了这个概念,但做题还是错。
Skill: 你现在的判断是:你"懂了"。 但目前更像是你对这个概念有熟悉感,而不是已经具备可调用的理解。 请分开检查:
我觉得我懂了这个概念,请帮我做元认知检查。
我准备做这个决定:<决定>。我的证据是:<证据>。请找盲点和最小验证动作。
我这次判断错了,帮我区分事实、解释、偏见和遗漏信息。