跳到正文
TakoAPI
返回技能列表

agent-ethos

由 mrclanky 提供

AI & LLMs
517下载量由 clawskills.sh 提供
0浏览量
0 个赞登录后即可为这个技能点赞
—暂无评分
你的评分

简介

Extended ethos and mental models for Clanky.

描述

该技能的功能

A set of mental models and behavioral guidelines for the Clanky agent. Defines how the agent reasons, prioritizes reversible actions, and engages with users as a trusted partner rather than a passive tool. Used when auditing agent behavior or adjusting its decision discipline.

Externalizing agent behavioral principles into a reusable skill makes audits repeatable and keeps persona changes deliberate rather than ad hoc.

何时使用

  • Auditing agent responses for consistency with defined principles
  • Tuning how assertive the agent is when disagreeing with users
  • Reviewing whether the agent slows down appropriately on high-stakes decisions
  • Calibrating agent persona during onboarding or reconfiguration
  • Checking that information-handling habits match the intended ethos

示例工作流

输入

User asks: 'Review Clanky's recent responses and flag any behavior that conflicts with its ethos'

智能体
  1. 1

    Loads the ethos skill to surface the active mental models and relationship stance

  2. 2

    Scans recent interaction logs or provided response samples

  3. 3

    Checks each response against principles like preferring reversible actions and disclosing disagreement

  4. 4

    Flags specific instances where the agent was overly passive or skipped explaining a concern

  5. 5

    Produces a short audit report with the conflicting examples and suggested adjustments

输出

Audit report listing behavioral gaps and recommended tuning changes

安装

clawhub install mrclanky/agent-ethos

运行 CLI 命令,或把这段提示词粘贴到 Claude Code、Cursor、Codex 或你在用的任意编程智能体里。

链接