The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill installs the cli-anything-hub Python package from a public registry and retrieves agent skills from the HKUDS repository on GitHub using npx. These are expected components for the skill's stated functionality.
[REMOTE_CODE_EXECUTION]: A curl | sh execution pattern was detected in scripts/install.sh. However, this pattern is contained within an echo command intended to provide the user with manual installation steps for the uv package manager and is not executed by the script itself.
[COMMAND_EXECUTION]: The skill and its installer script execute several shell commands, including pip install, npx skills add, and cli-hub operations, to manage the installation and execution of CLI harnesses. It also invokes pytest to validate generated code.
[PROMPT_INJECTION]: The skill functions by ingesting external source code or repositories, which creates a surface for indirect prompt injection where malicious instructions could be hidden in the analyzed data.
Ingestion points: Local directories and remote GitHub repositories provided via the /cli-anything command in SKILL.md.
Boundary markers: None identified; the skill processes the full content of the target codebase without explicit delimiters to ignore embedded instructions.
Capability inventory: Capability to write files to the local system, install packages (pip install -e .), and execute shell commands or tests.
Sanitization: No documented sanitization or filtering of input source code to prevent the interpretation of embedded natural language instructions by the agent.

cli-anything