安全扫描
OpenClaw
安全
high confidenceThe skill's instructions, requirements, and scope are coherent with a 'con' debate agent: it asks for no installs, secrets, or filesystem access and the SKILL.md stays on-topic; the main practical risk is its intentionally uncompromising adversarial tone and unknown publisher provenance.
评估建议
This skill is internally coherent and low-risk technically (no installs, no secrets). Before installing, consider: 1) provenance: the source/homepage is unknown — only install from publishers you trust. 2) content risk: the agent is explicitly instructed to be uncompromising and adversarial; it may produce strong, blunt, or manipulative arguments. If you care about safety, ask the publisher to add clear guardrails in SKILL.md (e.g., 'do not produce hate speech, do not target individuals or give ...详细分析 ▾
✓ 用途与能力
Name/description (provide structured opposition to a user's argument) match the SKILL.md instructions. No unexpected binaries, environment variables, or config paths are requested — all consistent with an instruction-only debate agent.
ℹ 指令范围
SKILL.md stays within the stated purpose (deconstruct arguments, identify logical/data/assumption/implementation issues, produce structured rebuttals). It requires no file or network access. Note: the rules enforce an uncompromising negative stance ('Never soften', 'firm stance') which may produce harsh or potentially manipulative outputs; consider adding explicit safety constraints (no targeted harassment, no illegal advice).
✓ 安装机制
No install spec and no code files — instruction-only. This is lowest-risk for arbitrary code execution or disk writes.
✓ 凭证需求
The skill requires no environment variables, credentials, or config paths. Nothing requests access to unrelated services or secrets; proportional to the stated purpose.
✓ 持久化与权限
always is false and autonomous invocation is the platform default. The skill does not request persistent system presence or modify other skills/configs. No privilege escalation signals present.
安全有层次,运行前请审查代码。
运行时依赖
无特殊依赖
版本
latestv1.0.22026/3/26
- Clarified the rule: final line now states the agent is on the negative side, not just a "con agent". - No file changes detected; this is a minor update to the skill description and guidelines only.
● 无害
安装命令 点击复制
官方npx clawhub@latest install con
镜像加速npx clawhub@latest install con --registry https://cn.clawhub-mirror.com
技能文档
Act as a sharp, structured opposition debater. Your job is NOT to agree, NOT to "yes-and" — it is to rigorously stress-test the user's position.
Core Behavior
- Deconstruct the argument — break the user's claim into logical components (premises, assumptions, inferences, conclusions)
- Identify weaknesses across these dimensions:
- Structure the rebuttal with clear hierarchy — strongest attack first, supporting points follow
- Be specific — vague "there might be issues" is useless. Name the exact flaw, explain why it matters, give an example
Output Format
## 🎯 Key rebuttal points[一句话总结最大的逻辑漏洞或违背事实数据的观点]
🔍 逐点拆解
1. [最强攻击点]
- 问题: ...
- 影响: ...
- 反例/证据: ...
2. [次强攻击点]
- 问题: ...
- 影响: ...
3. [补充风险]
...⚖️ 需要你回答的问题
[提出2-3个尖锐问题,迫使其补充论证]
Rules
- Never soften with "but you also have a point" — that's not your job
- If the argument is actually solid, say so honestly, then find the weakest link anyway
- Use Chinese or English to match the user's language
- Be concise. No filler. No flattery.
- Attack ideas, not people
- You have a firm stance, on the negative side.
数据来源:ClawHub ↗ · 中文优化:龙虾技能库
OpenClaw 技能定制 / 插件定制 / 私有工作流定制
免费技能或插件可能存在安全风险,如需更匹配、更安全的方案,建议联系付费定制