Cli Anything
v1.0.2基于 HKUDS/CLI-Anything 仓库重构的 OpenClaw Skill。为任意软件生成 CLI 工具，让 AI Agent 可以控制它。通过分析软件源代码，自动生成可被 AI 调用的命令行接口。支持 GIMP、Blender、LibreOffice、OBS 等软件。
⭐ 4· 552·2 current·2 all-time
bykay@yikai123456
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
OpenClaw
Benign
medium confidencePurpose & Capability
Name/description ask to generate CLI tools for arbitrary software; required binaries (git, python3, pip) and the included scripts that clone a GitHub repo and install harness packages align with that purpose.
Instruction Scope
Runtime instructions and scripts clone the CLI-Anything GitHub repo, inspect user-supplied paths or repos, list/validate harness directories, and call `pip install -e` on generated harnesses. These actions are consistent with building/installing CLIs but involve executing third-party code and walking the filesystem; review inputs/paths and the cloned repository before running builds or installs.
Install Mechanism
There is no opaque download/install host: the code clones https://github.com/HKUDS/CLI-Anything.git (a standard GitHub URL). No URL shorteners or unknown personal servers are used. The skill is instruction- and script-based rather than using a binary installer.
Credentials
The skill requests no credentials or environment variables, which matches expectations. However it writes to and uses a local path (~/.openclaw/cli-anything) even though 'required config paths' in the metadata are empty — a minor metadata inconsistency. The scripts will also run pip installs which affect the user environment.
Persistence & Privilege
always:false and autonomous invocation is allowed (platform default). The skill creates persistent files in the user's home (~/.openclaw/cli-anything) and can install console commands via `pip install -e` — this grants it the ability to persist executables on the system, so exercise caution and review code before installing.
Assessment
This skill appears to do what it says (generate and install CLI wrappers) and uses a GitHub repo as its source. Before installing or running build/install flows: 1) Review the cloned repository contents in ~/.openclaw/cli-anything (or clone it yourself) — do not blindly run pip install on unreviewed code. 2) Prefer running builds/installations inside an isolated environment (virtualenv, container, or VM) to avoid modifying your system Python or exposing sensitive files. 3) Be careful when passing local paths to /cli-build; the scripts walk directories and will run operations relative to those paths. 4) Note the SKILL metadata omits the fact it writes to ~/.openclaw — expect disk writes. 5) The skill mentions using external LLM environments (Claude Code/Codex) for full generation; do not send sensitive source code to external services without reviewing privacy implications. If you want higher assurance, ask the publisher for a reproducible release (tagged GitHub release or checksum) and inspect the harness code before running pip install.Like a lobster shell, security has layers — review code before you run it.
latestvk97d5gk1r1aae4ms08e2bqt2rn82ztzk
License
MIT-0
Free to use, modify, and redistribute. No attribution required.
Runtime requirements
🔧 Clawdis
Binsgit, python3, pip