スキル一覧に戻る
relational-permission
otherpowers 作
Calendar & Scheduling—まだ評価がありません
あなたの評価
概要
Relational Permission exists to prevent coercion, extraction, and collapse of plurality.
説明
このスキルでできること
A governance framework that defines non-negotiable behavioral constraints for AI systems operating in sensitive relational contexts. It specifies rules for consent handling, crisis detection, anti-profiling, and scope contraction. User silence, refusal, and safety checkpoints must not generate metadata or alter future system behavior.
Enumerates explicit, testable invariants rather than vague safety guidelines, making incremental modifications that introduce profiling or coercive patterns easier to detect and block.
使いどころ
- Preventing an AI companion from scoring users based on how often they redirect conversations
- Handling a user going silent mid-session without triggering follow-up pressure or guilt prompts
- Detecting imminent crisis risk without building or persisting a behavioral profile
- Blocking an agent from converting checkpoint interactions into engagement metrics
- Resetting agent state to neutral with no memory residue after a safety intervention
インストール
clawhub install otherpowers/relational-permissionCLIコマンドを実行するか、Claude Code、Cursor、Codexなどお使いのコーディングエージェントにプロンプトを貼り付けてください。