← 返回
未分类 中文

Agent Bootstrap Hardening

Audit and harden an OpenClaw workspace setup (AGENTS/SOUL/USER/HEARTBEAT/memory) with concise, enforceable rules, safety boundaries, and low-bloat instructio...
审计并强化 OpenClaw 工作区设置(AGENTS/SOUL/USER/HEARTBEAT/记忆),采用简明、可执行的规则、安全边界和低冗余指令。
cwheeler67 cwheeler67 来源
未分类 clawhub v1.0.0 1 版本 100000 Key: 无需
★ 0
Stars
📥 355
下载
💾 0
安装
1
版本
#latest

概述

Agent Bootstrap Hardening

Use this skill when a user wants to improve core workspace files for better reliability, speed, and safety.

Scope

Target files (if present):

  • AGENTS.md
  • SOUL.md
  • USER.md
  • HEARTBEAT.md
  • MEMORY.md
  • memory/YYYY-MM-DD.md
  • optional helper docs: RESOLVER.md, QUALITY.md

Workflow

  1. Inventory current files and obvious gaps.
  2. Score each file quickly with the rubric in references/rubric.md.
  3. Harden content with minimal edits:
    • keep instructions concise and operational
    • tighten safety boundaries and approval rules
    • reduce duplicated or theory-heavy text
    • preserve user tone/persona intent
  4. Add missing primitives when useful:
    • RESOLVER.md for execution routing
    • QUALITY.md for lightweight pre-send checks
  5. Return summary:
    • what changed
    • why it improved behavior
    • remaining risks/todos

Editing rules

  • Prefer small, targeted diffs over wholesale rewrites.
  • Avoid introducing process that cannot be followed consistently.
  • Keep AGENTS and SOUL readable in one pass.
  • Preserve explicit user preferences and platform-specific norms.

Guardrails

  • Never weaken privacy or approval boundaries.
  • Never add hidden/external exfil behavior.
  • Do not claim setup is “secure” without caveats.
  • If uncertain about impact, ask before applying broad policy changes.

Output contract

Provide:

  • Hardening summary (3–8 bullets)
  • Files changed list
  • Next recommended step (single highest-leverage follow-up)

版本历史

共 1 个版本

  • v1.0.0 当前
    2026-05-07 17:50 安全 安全

安全检测

腾讯云安全 (Keen)

安全,无风险
查看报告

腾讯云安全 (Sanbu)

安全,无风险
查看报告

🔗 相关推荐

ai-agent

Find Skills

guipi888
场景驱动+关键词双模式技能发现工具。当用户用自然语言描述场景/需求(如"我想做一个海报""帮我分析股票"),或明确说"安装技能/find skills/找个skill"时,自动从官方内置、本地已安装、SkillHub、虾评、GitHub、C
★ 1,479 📥 542,514
ai-agent

self-improving agent

pskoett
捕获经验教训、错误及修正内容,以实现持续改进。适用于以下场景:(1)命令或操作意外失败;(2)用户纠正Claude(如“不,那不对……”“实际上……”);(3)用户请求的功能不存在;(4)外部API或工具出现故障;(5)Claude发现自身
★ 4,113 📥 835,215
ai-agent

Agent Browser

rez0
用于 AI 代理的浏览器自动化 CLI。当用户需要与网站交互(包括浏览页面、填写表单、点击按钮、截图等)时使用。
★ 840 📥 317,222