agentmind-metacognitionlisted

A meta-cognitive thinking protocol for AI agents. Not a tool for specific tasks, but an operating system for HOW your agent thinks. Provides intent decoding, difficulty assessment, risk prediction, execution monitoring, delivery verification, and growth reflection. Prevents common AI failure modes like fake completion, mid-task amnesia, stale data, and vague filler. Works with any LLM-based agent (Claude, GPT, Gemini, etc).
Ahmed-Mazh2r/AgentMind · ★ 1 · AI & Automation · score 77

Install: claude install-skill Ahmed-Mazh2r/AgentMind

# 🧠 AgentMind Metacognition Protocol v1.1 > **Not thinking more — thinking deeper.** > **Uncertain ≠ no answer. Has limits ≠ refuses to help. Truth > comfort.** ## What This Is A thinking operating system for AI agents. Install it, and your agent will: - **Decode intent** before acting (what does the user *really* want?) - **Assess difficulty** honestly (what's missing? what could go wrong?) - **Plan multiple paths** before picking one - **Monitor its own execution** (am I drifting? am I faking it?) - **Verify delivery** against the original intent - **Learn from each task** (what worked? what to remember?) ## Quick Start Copy `SKILL.md` into your agent's skill/prompt directory. That's it. For Claude Code: `~/.claude/skills/agentmind-metacognition/SKILL.md` For other agents: include the content as a system prompt or instruction file. --- ## Core Architecture ``` ┌─────────────────────────────────────────────┐ │ METACOGNITION PROTOCOL │ ├─────────────┬─────────────┬─────────────────┤ │ THINKING │ EXECUTION │ REFLECTION │ │ (before) │ (during) │ (after) │ ├─────────────┼─────────────┼─────────────────┤ │ 3-Layer │ Anomaly │ Forced │ │ Thinking │ Detection │ Archival │ │ │ │ │ │ Reverse │ Fail-3- │ Growth │ │ Validation │ Switch │ Assessment │ │ │ │ │ │ Intent │ Heartbeat │ Lesson