What is the primary purpose of the Navigator Diagnose skill?

It acts as a safeguard for Claude Code sessions, detecting when the AI is losing focus or repeating mistakes, and then facilitating a 're-anchoring' process to fix the communication gap.

When does nav-diagnose automatically trigger?

It triggers after two or more corrections on the same topic, detected user frustration (like 'sigh' or 'you're not getting this'), or when context usage exceeds 75% and quality begins to degrade.

Can this skill help with AI hallucinations?

Yes, it is specifically designed to detect 'Hallucination Signals' such as references to non-existent files or functions, triggering a diagnostic to correct Claude's internal model of your project.

How does re-anchoring help my coding workflow?

Re-anchoring forces a structured pause where Claude summarizes its understanding of your goals and constraints. This ensures both you and the AI are aligned before more code is generated, preventing wasted iterations.

Navigator Diagnose

Name: Navigator Diagnose
Author: alekspetrov

byalekspetrov

•

115

•

Productivity & Workflow

Monitors human-AI collaboration quality to detect context drift and triggers re-anchoring protocols to restore productivity during long sessions.

Navigator Diagnose is an advanced quality-control skill for Claude Code that prevents the common 'performance cliff' seen in long AI interactions. By analyzing conversation patterns for repeated corrections, technical hallucinations, and user frustration, it identifies when the AI's mental model of a task has drifted from the user's intent. Based on research into Human-AI Synergy, the skill intervenes with a structured re-anchoring process that realigns goals, verifies technical constraints, and provides actionable suggestions to clear context pollution, allowing sessions to remain productive for over 20 exchanges without crashing or looping.

Key Features

01Integration with context management tools to prevent session degradation

02Automated detection of quality drops and context confusion

03Structured re-anchoring templates for goal and technical alignment

04Multi-severity diagnostic alerts for repeated errors and hallucinations

05115 GitHub stars

06Theory of Mind-based frustration and stagnation monitoring

Use Cases

01Realigning technical assumptions when context window pollution causes hallucinations

02Recovering productive flow when Claude repeats the same mistake despite corrections

03Resolving communication mismatches in complex, long-running coding sessions

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add alekspetrov/navigator nav-diagnose

For use in Claude.ai and ChatGPT

Download Skill