← 返回 Skills 市场
134
总下载
0
收藏
0
当前安装
1
版本数
在 OpenClaw 中安装
/install tool-output-auditor
功能描述
Audit tool and command results before acting on them. Use when an agent has already run a command or tool, especially in multi-step workflows where a failed,...
安全使用建议
This skill appears safe and coherent for auditing command output: it runs locally with python3, reads only the file or stdin you give it, and prints a verdict. Before using, consider: (1) do not feed it sensitive secrets or live logs you don't want processed; the script only classifies text you provide; (2) review/adjust the regex patterns if your workflows use unusual success/failure messages (false positives/negatives are possible); (3) it's not networked and does not exfiltrate data, but treat any tool that reads files/stdin with the usual caution—only pass outputs you intend to analyze. Overall, the skill matches its description and asks for nothing disproportionate.
功能分析
Type: OpenClaw Skill
Name: tool-output-auditor
Version: 0.1.0
The skill is a defensive utility designed to help AI agents accurately interpret command outputs and avoid hallucinating success in multi-step workflows. It uses a Python script (scripts/audit_output.py) to perform regex-based classification of text into success, failure, or partial states. The documentation and code are transparent, lack high-risk behaviors, and focus entirely on improving agent reliability and validation.
能力评估
Purpose & Capability
Name/description match the provided assets: SKILL.md explains auditing tool output and the bundled scripts (scripts/audit_output.py plus reference docs) implement that purpose. Requiring python3 is appropriate and proportional.
Instruction Scope
Runtime instructions limit behavior to reading command output (from a file or stdin) and producing a classification and recommendation. There are no instructions to read unrelated system files, access environment secrets, or transmit data externally.
Install Mechanism
No install spec is present (instruction-only with a bundled script). Nothing is downloaded or extracted; the only runtime requirement is an existing python3 binary.
Credentials
The skill declares no required env vars, credentials, or config paths. The code does not reference environment secrets or external APIs—only local stdin/file input and pattern matching.
Persistence & Privilege
always is false, the skill is user-invocable and not forced into all agent runs. It does not modify other skills or request persistent system presence.
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 tool-output-auditor - 安装完成后，直接呼叫该 Skill 的名称或使用
/tool-output-auditor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
v0.1.0
Initial release of the Tool Output Auditor skill.
- Provides guidelines for auditing tool and command outputs before proceeding in workflows.
- Introduces a 4-state classification for outputs: success, failure, partial, or ambiguous.
- Describes common pitfalls and anti-patterns when handling tool outputs.
- Includes scripts/audit_output.py to classify output files and suggest next actions.
- Offers references and checklists for safer workflow automation.
元数据
常见问题
Tool Output Auditor 是什么？
Audit tool and command results before acting on them. Use when an agent has already run a command or tool, especially in multi-step workflows where a failed,... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 134 次。
如何安装 Tool Output Auditor？
在 OpenClaw 或 Claude Code 对话框中运行命令「/install tool-output-auditor」即可一键安装，无需额外配置。
Tool Output Auditor 是免费的吗？
是的，Tool Output Auditor 完全免费，采用 MIT-0 许可证，可自由下载、安装和使用。
Tool Output Auditor 支持哪些平台？
Tool Output Auditor 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 Tool Output Auditor？
由 wuu Dao（@daowuu）开发并维护，当前版本 v0.1.0。
推荐 Skills