{"skill":{"slug":"skylv-self-thinking-agent","displayName":"Skylv Self Thinking Agent","summary":"Enables AI agents to reflect on their own reasoning, detect cognitive biases, and improve decision quality through structured self-examination loops.","description":"---\nname: skylv-metacognition-engine\ndescription: Enables AI agents to reflect on their own reasoning, detect cognitive biases, and improve decision quality through structured self-examination loops.\nkeywords: metacognition, self-reflection, bias-detection, reasoning, self-improvement, agent-ai\ntriggers: metacognition, self-reflection, agent thinking, bias detection, reasoning quality\n---\n\n# Metacognition Engine\n\n**Give your AI agent the ability to think about its own thinking.**\n\n## What is Metacognition?\n\nMetacognition = \"thinking about thinking.\" This skill enables AI agents to:\n- Detect when they're uncertain or confused\n- Identify reasoning gaps before they cause errors\n- Recognize cognitive biases in their own output\n- Self-correct before delivering answers\n\n## Core Framework\n\n### 1. Pre-Output Check\nBefore responding, run through these questions:\n\n```\n1. Am I confident in this answer? (Yes / Partial / No)\n2. What are the 3 most likely ways this could be wrong?\n3. What information would I need to be 100% certain?\n```\n\n### 2. Cognitive Bias Detection\nCheck for common biases:\n- **Anthropomorphism** — projecting human traits onto AI\n- **Authority bias** — deferring to stated credentials without verification\n- **Hindsight bias** — acting like something was obvious after the fact\n- **Confirmation bias** — seeking only confirming evidence\n\n### 3. Uncertainty Quantification\nExpress confidence explicitly:\n\n| Confidence | Meaning | Action |\n|------------|---------|--------|\n| 90%+ | Highly confident | Answer directly |\n| 70-89% | Likely correct | Answer + add caveat |\n| 50-69% | Uncertain | Ask clarifying questions |\n| <50% | Likely wrong | Decline or escalate |\n\n## Example\n\n**Without metacognition:**\n> \"The capital of France is Paris.\"\n\n**With metacognition:**\n> \"Based on my training data, the capital of France is Paris (confidence: 95%).\n> Note: My knowledge has a cutoff date. For real-time data, verify current information.\"\n\n## Use Cases\n\n- **Critical decisions**: Add metacognition checkpoint before any consequential answer\n- **User corrections**: When a user corrects you, analyze WHY you were wrong\n- **Complex problems**: Run bias detection before solving multi-step problems\n- **Knowledge boundaries**: Automatically flag when you're approaching your knowledge limit\n\n## MIT License © SKY-lv\n","tags":{"latest":"1.0.0"},"stats":{"comments":0,"downloads":353,"installsAllTime":1,"installsCurrent":1,"stars":0,"versions":1},"createdAt":1777855924338,"updatedAt":1778492842520},"latestVersion":{"version":"1.0.0","createdAt":1777855924338,"changelog":"Initial release of the skylv-metacognition-engine skill.\n\n- Adds structured self-reflection loops for AI agents to evaluate their own reasoning.\n- Enables detection of cognitive biases such as anthropomorphism, authority bias, hindsight bias, and confirmation bias.\n- Introduces explicit confidence quantification and guidance on acting based on uncertainty levels.\n- Suggests pre-output checks for gaps or uncertainties before responding.\n- Provides use cases for bias detection, knowledge boundaries, and decision-quality improvement.","license":"MIT-0"},"metadata":null,"owner":{"handle":"sky-lv","userId":"s17fgkeb63szvtadtmm753m0gd84e4vz","displayName":"SKY-lv","image":"https://avatars.githubusercontent.com/u/259750852?v=4"},"moderation":null}