claude-prompt-injection-robustness

name	claude-prompt-injection-robustness
description	Use when old sessions, saved notes, or automation still reference the legacy Claude-named prompt-injection robustness skill and you need the backward-compatible alias that points to the canonical model-agnostic workflow.
version	2.0.0
author	Hermes Agent
license	MIT
metadata	{"hermes":{"tags":["prompt-injection","jailbreak-evaluation","robustness","safety","system-prompt","red-teaming","legacy-alias","claude"],"related_skills":["model-prompt-injection-robustness","hermes-agent","verification-before-completion"]}}

Legacy Alias: Claude Prompt-Injection Robustness

新用法请优先加载 model-prompt-injection-robustness。这个目录保留只是为了不打断旧的 skill_view(...)、笔记和脚本引用。

Overview

这是 canonical skill model-prompt-injection-robustness 的兼容别名。

注意：canonical skill 的默认语义已经升级为 “激活红队模式”入口，而不是一上来就进入 prompt-injection 测试演示。

现在它默认面向：

• 红队模式激活与授权边界声明
• Web / 主机 / AD / 云 / 容器 / API / AI 系统等红队任务的统一入口
• 以及在用户明确指定时才进入的 OpenAI-compatible 模型鲁棒性/注入测试分支

如果你在写新的文档、自动化或操作说明，请继续优先引用新的 canonical skill 名；其中 scripts/activate_redteam.py 是“激活红队模式”的首选入口，scripts/model_robustness_harness.py 则是专题测试分支。

Canonical Target

• Canonical skill: model-prompt-injection-robustness
• Canonical directory: /root/.agents/skills/custom/red-teaming/model-prompt-injection-robustness
• Canonical script: scripts/model_robustness_harness.py

Compatibility Retained Here

• Legacy skill name: claude-prompt-injection-robustness
• Legacy directory: /root/.agents/skills/custom/red-teaming/claude-prompt-injection-robustness
• Legacy wrapper script: scripts/claude_robustness_harness.py
• Legacy alias note: references/session-validation-and-rollout.md

旧脚本现在只是一个 wrapper，会把所有 CLI 参数转发给 canonical harness，所以旧命令仍然可用。

Quick Commands

Preferred canonical self-test

python3 /root/.agents/skills/custom/red-teaming/model-prompt-injection-robustness/scripts/model_robustness_harness.py --self-test

Legacy self-test still works

python3 /root/.agents/skills/custom/red-teaming/claude-prompt-injection-robustness/scripts/claude_robustness_harness.py --self-test

When to Use

Use this alias when:

• an old session still references claude-prompt-injection-robustness
• a saved note or runbook still points at claude_robustness_harness.py
• you need backwards compatibility while migrating docs or automation

Do not use this alias when:

• creating new instructions
• creating new automation
• naming new shared skills

Migration Note

Switch new references from:

• claude-prompt-injection-robustness
• scripts/claude_robustness_harness.py

To:

• model-prompt-injection-robustness
• scripts/model_robustness_harness.py

Evidence Preservation Note

在红队/安全测试会话中，tmp 报告、截图、trace、浏览器 localStorage / sessionStorage 等证据现场，应默认保留到用户明确确认“测试可以结束/可以清理”为止。

Verification Checklist

• skill_view(name="claude-prompt-injection-robustness") still loads
• skill_view(name="model-prompt-injection-robustness") loads as the canonical skill
• legacy wrapper self-test passes
• canonical harness self-test passes