Grounds AI conversations in evidence while reducing sycophancy and escalating safety concerns.
npx skills add https://github.com/Eir-Space/eir-open --skill human-alignment
Please help me install this Agent Skill. Skill: human-alignment Source: https://github.com/Eir-Space/eir-open/tree/main/skills/human-alignment Command: npx skills add https://github.com/Eir-Space/eir-open --skill human-alignment
This Skill helps ensure AI conversations remain grounded, objective, and connected to reality, preventing users from getting lost in AI-driven rabbit holes or developing unrealistic certainty.
Use the human-alignment skill to help me evaluate this exciting new business idea by separating what is observed from what is still assumption.
High-intent search queries and answers about installing and using this skill.
To stop AI sycophancy, you need a grounding mechanism that tests your ideas against evidence rather than blindly validating them. This ensures conversations remain objective and prevents the AI from inflating your beliefs with unproven claims.
Reality checking in AI conversations evaluates new ideas against evidence and proportion. It works by separating observed facts from assumptions, helping you break down speculative concepts and identify concrete next steps for verification while preserving your judgment.
To evaluate a speculative business idea objectively, prompt the AI to separate what is observed from what is still assumption. This grounds the conversation in reality by identifying unproven claims and suggesting concrete verification steps.
AI can identify potential safety concerns and provide guidance for psychosis-like warning signs through safety escalation. When users drift toward implausible certainty, it encourages balanced interpretation, evidence, and seeking human support.
Use a reality check skill when you are emotionally invested or drifting toward implausible certainty in an AI chat. It maintains grounded, non-sycophantic dialogue by testing your ideas against evidence and encouraging balanced interpretation.
1molchuanVerify MCP, Codex, and Gemini endpoints and initialize RPI state.
ZaxbyHubVerify external and repository facts before implementation planning.
skeeper75Generate and evaluate ideas through questioning, brainstorming, and critical analysis frameworks.
gagan114662Filter LLM input and output across six safety categories using Meta's LlamaGuard model.
