Can I see when an escalation happens?

Yes, the skill uses the Task tool to spawn a deep-think-partner, which is visible in the execution flow, ensuring transparency in the reasoning process.

Does this help with 'hallucinations'?

Yes, by detecting uncertainty and forcing a more structured analysis phase, it significantly reduces the likelihood of the AI providing non-committal or incorrect answers to difficult questions.

How does this skill interact with deep-think-partner?

When complexity thresholds are met, the skill triggers a subagent task using a detailed prompt template to analyze the problem deeply before providing the user with a synthesized recommendation.

What is a struggle signal in AI?

A struggle signal is a pattern in AI reasoning—such as hedging, verbosity, or mid-response contradiction—that indicates the model is reaching the limits of its immediate reasoning capacity for a specific task.

Metacognitive Guard

Name: Metacognitive Guard
Author: ANcpLua

byANcpLua

•

生産性とワークフロー

Monitors Claude's reasoning processes to detect struggle signals and proactively escalate complex tasks to specialized deep-thinking agents.

概要

Metacognitive Guard is a specialized Claude Code skill designed to enhance AI reliability by monitoring for internal struggle signals such as hedging, contradiction, or circular reasoning. It provides a structured framework for the AI to self-assess its confidence levels across architectural decisions, ambiguous requirements, and multi-domain synthesis. By integrating with deep-thinking resources, it ensures that high-stakes or high-complexity problems receive the necessary cognitive depth before a response is generated, resulting in more decisive, accurate, and actionable recommendations for the developer.

主な機能

Automated escalation workflows to deep-think-partner agents
Automatic detection of struggle signals like hedging and verbosity
Performance metrics for tracking escalation accuracy and outcomes
1 GitHub stars
Structured synthesis of deep-thinking results for user delivery
Proactive self-assessment checklist for high-complexity tasks

ユースケース

Resolving highly ambiguous project requirements with significant rework costs
Navigating complex architectural trade-offs between competing technologies
Systematic analysis of edge cases and failure modes in concurrent systems

概要

主な機能

Automated escalation workflows to deep-think-partner agents
Automatic detection of struggle signals like hedging and verbosity
Performance metrics for tracking escalation accuracy and outcomes
1 GitHub stars
Structured synthesis of deep-thinking results for user delivery
Proactive self-assessment checklist for high-complexity tasks

ユースケース

Resolving highly ambiguous project requirements with significant rework costs
Navigating complex architectural trade-offs between competing technologies
Systematic analysis of edge cases and failure modes in concurrent systems