Can I see the internal scores of the Proprioception engine?

Yes, you can request a 'proprioception dashboard' at any time to see the scores for Goal Proximity, Confidence, Drift, Capability, and Session Quality.

Does this skill work automatically in the background?

Yes, once installed, it runs silently and only intervenes or alerts the user if a specific threshold is breached, such as high goal drift or low response quality.

What is Proprioception in the context of an AI agent?

Proprioception is the ability for an AI agent to sense its own 'position' relative to a user's goal, its own confidence boundaries, and the overall trajectory of the conversation.

How does Proprioception help prevent AI hallucinations?

It uses Capability Boundary Sensing and Confidence Topography to identify when the agent is moving from factual ground into speculative territory, prompting the agent to disclose uncertainty.

Does this skill require external API calls or extra costs?

No, Proprioception uses lightweight mathematical analysis and heuristics on the conversation text locally, meaning it has zero external API costs and minimal overhead.

Proprioception AI Agent Sense

Name: Proprioception AI Agent Sense
Author: dvcrn

bydvcrn

•

분석 및 모니터링

Enhances AI agents with self-spatial awareness to monitor goal alignment, confidence levels, and conversation quality in real-time.

Proprioception equips Claude Code agents with a continuous, real-time awareness of their own performance and alignment with user objectives. By monitoring five key dimensions—goal proximity, confidence mapping, drift detection, capability boundaries, and session quality—this skill allows the agent to recognize when it is straying off-course or nearing the limits of its knowledge. It functions as a lightweight, local monitoring layer that improves reliability and prevents unproductive conversation loops without requiring external API calls or database overhead, acting as an essential safety and quality-control mechanism for complex tasks.

주요 기능

01Capability Boundary Sensing (CBS) to alert the user when the agent reaches its knowledge limits

02Session Quality Pulse (SQP) to track the cumulative health and utility of the interaction turns

03Goal Proximity Radar (GPR) to ensure responses remain aligned with the user's original intent

04Drift Detection (DD) to identify and correct circular, tangential, or degenerative conversation patterns

05Confidence Topography (CT) to map factual certainty versus speculative territory in outputs

063 GitHub stars

사용 사례

01Preventing hallucinations during complex technical discussions by mapping confidence zones

02Maintaining focus on long-term project objectives during multi-turn development sessions

03Identifying and resetting unproductive conversation loops that waste user time and tokens

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add dvcrn/openclaw-skills-marketplace proprioception

For use in Claude.ai and ChatGPT

주요 기능

01Capability Boundary Sensing (CBS) to alert the user when the agent reaches its knowledge limits

02Session Quality Pulse (SQP) to track the cumulative health and utility of the interaction turns

03Goal Proximity Radar (GPR) to ensure responses remain aligned with the user's original intent

04Drift Detection (DD) to identify and correct circular, tangential, or degenerative conversation patterns

05Confidence Topography (CT) to map factual certainty versus speculative territory in outputs

063 GitHub stars

사용 사례

01Preventing hallucinations during complex technical discussions by mapping confidence zones

02Maintaining focus on long-term project objectives during multi-turn development sessions

03Identifying and resetting unproductive conversation loops that waste user time and tokens