The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill (SKILL.md, references/workflow.md) instructs users to download and install the python-cli-demo package from PyPI and an unverified GitHub repository (github.com/username/python-cli-demo), which lacks verifiable ownership or trust.
[REMOTE_CODE_EXECUTION]: The instructions facilitate remote code execution by providing commands to install software directly from an external Git repository (pip install git+https://github.com/username/python-cli-demo.git).
[COMMAND_EXECUTION]: In the troubleshooting section of references/workflow.md, the skill recommends using sudo for package installation on Linux/macOS. This encourages users to grant elevated permissions to an external installation process, which is a significant security risk.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it processes user-provided input via CLI arguments (e.g., demo-cli hello [name]) and executes them through subprocess.run without sanitization.
Ingestion points: Command line arguments for demo-cli documented in SKILL.md and references/workflow.md.
Boundary markers: None present in the provided examples.
Capability inventory: The skill uses subprocess.run to execute commands on the host system.
Sanitization: No input validation or escaping logic is included in the implementation examples.

demo-cli