← 返回 Skills 市场

CLI-Anything for OpenClaw

Name: CLI-Anything for OpenClaw
Author: barrontang

作者 Barron Tang · GitHub ↗ · v0.1.0 · MIT-0

cross-platform ⚠ suspicious

1530

总下载

当前安装

版本数

在 OpenClaw 中安装

/install cli-anything-openclaw

功能描述

Adapt HKUDS CLI-Anything for OpenClaw workflows. Use when the user wants to build, refine, test, or validate an agent-native CLI harness for a GUI applicatio...

使用说明 (SKILL.md)

Use this skill as an OpenClaw-native bridge to the upstream CLI-Anything methodology.

What this skill is

This skill packages the core CLI-Anything methodology in a portable form for OpenClaw users.

Use it when the user wants to:

analyze whether a software project is a good fit for an agent-friendly CLI harness
design command groups and workflow structure for a GUI application or repo
refine an existing harness
validate packaging, tests, and CLI entry points using CLI-Anything-style thinking

Inputs

Accept either:

a local source path
a GitHub repository URL

Derive the software name from the local directory name if needed.

Required references

Read these in order:

references/harness.md — core methodology and workflow
references/codex-skill.md — condensed Codex-oriented rules
Relevant command references as needed:
- references/cli-anything.md
- references/cli-anything-refine.md
- references/cli-anything-validate.md
- references/cli-anything-test.md
- references/cli-anything-list.md

Preferred workflow

Acquire source locally if needed.
Analyze architecture, backend surfaces, data model, and current automation affordances.
Decide whether the project is a good fit for an agent-native CLI harness.
If building:
- propose command groups
- propose state model
- propose agent-harness/ layout
If refining:
- inventory current commands and tests
- identify gaps
- prioritize high-impact additions
If validating:
- check packaging, CLI namespace, tests, and JSON output expectations
Always report:
- files changed or proposed
- commands run
- risks / limitations
- next best step

Important constraints

Treat this as a methodology skill, not as a promise that a harness already exists.
Prefer wrapping real software backends rather than reimplementing behavior.
Plan tests before writing them.
Keep outputs concrete and actionable.

安全使用建议

This skill is a coherent methodology pack: it tells the agent how to analyze, build, refine, test, and validate CLI harnesses for GUI apps. Before installing or running it, be aware that using the skill will typically cause the agent to: clone or read source repositories you provide, inspect local files (setup.py, tests, HARNESS.md), run test suites (pytest), and invoke local/external executables found on the host (via subprocess). Those actions can modify files, create artifacts, or trigger installed software; they do not require any secrets from you. If you do not fully trust the target repository or want to limit risk, run the skill in a sandboxed environment or provide a read-only copy of the repo and review proposed command lists/changes before allowing execution.

功能分析

Type: OpenClaw Skill Name: cli-anything-openclaw Version: 0.1.0 The skill bundle implements the CLI-Anything methodology, which directs an AI agent to clone GitHub repositories, generate Python code, and execute system commands via subprocesses and 'pip install'. While these actions are consistent with the stated goal of building CLI harnesses for GUI applications (as detailed in SKILL.md and references/harness.md), they grant the agent broad, high-risk capabilities including arbitrary code execution and file system modification. No clear evidence of intentional malice, data exfiltration, or obfuscation was found, but the inherent risk of the automated 'build-and-test' workflow qualifies it as suspicious.

能力评估

✓ Purpose & Capability

The name/description and the SKILL.md consistently describe a methodology for building/refining/testing CLI harnesses. No unrelated environment variables, binaries, or install steps are required, which is proportional for an instruction-only methodology skill.

ℹ Instruction Scope

The runtime instructions direct the agent to clone repositories, read source files (setup.py, HARNESS.md, tests, etc.), run pytest, and invoke local backends via subprocess (shutil.which/subprocess.run). This is coherent for the stated purpose but means the agent will access local filesystem and may execute local binaries when invoked — the user should be aware that the agent may run commands and test suites against the provided codebase.

✓ Install Mechanism

No install spec and no code files to execute are provided by the skill. That lowers disk-write and supply-chain risk; all behavior is driven by textual instructions and references packaged with the skill.

✓ Credentials

The skill does not request environment variables, credentials, or config paths. It does mention an optional test-time env var (CLI_ANYTHING_FORCE_INSTALLED) in the reference docs but does not require secrets; the env usage is justified by the testing workflow.

✓ Persistence & Privilege

always is false and there is no instruction to modify other skills or global agent settings. The skill is user-invocable and can be run autonomously by the agent (platform default), which is expected for this type of methodology skill.

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 cli-anything-openclaw
安装完成后，直接呼叫该 Skill 的名称或使用 /cli-anything-openclaw 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 public release.

元数据

Slug cli-anything-openclaw

版本 0.1.0

许可证 MIT-0

累计安装 13

当前安装数 13

历史版本数 1

常见问题