← 返回 Skills 市场
124
总下载
1
收藏
0
当前安装
1
版本数
在 OpenClaw 中安装
/install self-correction
功能描述
模型自我检查与纠错技能。当用户表达反对意见或指出错误时触发，如"不对"、"你回答错了"、"不是这样"、"重新理解"、"等等"、"我说的不是这个意思"等表达。用于重新分析和优化用户原始意图，而非修改上一轮回答。
安全使用建议
This skill appears coherent and limited to conversational behavior: it will detect user disagreement phrases, re-interpret the user's original prompt, and produce a revised answer. Before installing, confirm you are comfortable with an autonomous skill having permission to be invoked by the agent (the default) and review the SKILL.md examples to ensure the confirmation/clarification style matches your expectations. The included scripts are simple local packaging/validation helpers — no network downloads or secret usage were found.
功能分析
Type: OpenClaw Skill
Name: self-correction
Version: 1.0.0
The skill bundle implements a 'self-correction' mechanism designed to help an AI agent re-evaluate user intent when a mistake is pointed out. The SKILL.md file contains detailed instructions for handling user feedback and includes cryptographic signatures (ReservedCode) likely used for content provenance by the producer (Minimax). The accompanying Python scripts in the scripts/ directory are standard utility tools for validating and packaging the skill, and they do not exhibit any malicious behavior such as data exfiltration or unauthorized execution.
能力标签
能力评估
Purpose & Capability
The name and description describe a self-correction / intent-recovery helper, and the SKILL.md, _meta.json, and examples all implement that behavior. Required capabilities (none) are proportional to the stated purpose.
Instruction Scope
Runtime instructions are limited to dialog-level actions (detect trigger phrases, re-examine the user's previous input, propose revised interpretations and answers). The SKILL.md does not instruct the agent to read system files, access unrelated environment variables, call external endpoints, or exfiltrate data.
Install Mechanism
There is no install spec; the skill is instruction-only. The two included scripts are local packaging/validation utilities and do not download or execute remote code.
Credentials
The skill requests no environment variables, credentials, or config paths. The behavior described does not require additional secrets or cloud access.
Persistence & Privilege
Flags show always:false and default autonomous invocation allowed (normal). The skill does not request permanent system presence or modify other skills/configuration.
如何使用
- 确保已安装 OpenClaw（本地或 Docker 部署）
- 在对话框中输入安装命令：
/install self-correction - 安装完成后，直接呼叫该 Skill 的名称或使用
/self-correction触发 - 根据 Skill 的参数说明提供必要输入，即可获得结构化输出
版本历史
v1.0.0
self-correction 1.0.0
- Initial release providing model self-check and self-correction capability.
- Detects user denials, objections, or correction cues in Chinese and triggers a re-analysis of the original user intent.
- Explicitly avoids editing or apologizing for previous answers; instead, focuses on understanding and re-optimizing the user’s original prompt.
- Offers structured output: self-check alert, re-understood requirements, problem point analysis, and a new answer.
- Includes extensive detection templates, typical scenarios, and a checklist to ensure outputs are focused, concise, and non-repetitive.
元数据
常见问题
self-correction 是什么？
模型自我检查与纠错技能。当用户表达反对意见或指出错误时触发，如"不对"、"你回答错了"、"不是这样"、"重新理解"、"等等"、"我说的不是这个意思"等表达。用于重新分析和优化用户原始意图，而非修改上一轮回答。 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 124 次。
如何安装 self-correction？
在 OpenClaw 或 Claude Code 对话框中运行命令「/install self-correction」即可一键安装，无需额外配置。
self-correction 是免费的吗？
是的，self-correction 完全免费，采用 MIT-0 许可证，可自由下载、安装和使用。
self-correction 支持哪些平台？
self-correction 跨平台运行，可在任意部署了 OpenClaw / Claude Code 的环境中使用（cross-platform）。
谁开发了 self-correction？
由 华胥（@hauxu）开发并维护，当前版本 v1.0.0。
推荐 Skills