The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill uses https://ghfast.top as a mirror for cloning repositories in scripts/detect_network.py and references/network-mirrors.md. This domain has been identified as malicious by automated scanners and is not a verified or trusted source for code distribution.
[COMMAND_EXECUTION]: Several scripts, including scripts/collect_env_info.py and scripts/detect_network.py, use subprocess.run to execute shell commands for system discovery. The SKILL.md also instructs the agent to use docker exec for various installation and configuration tasks.
[REMOTE_CODE_EXECUTION]: Clones and installs packages from external sources (including an untrusted mirror) using pip install -e ., which allows the execution of arbitrary code during the installation process through setup scripts.
[DATA_EXFILTRATION]: Collects environment information like GPU details, glibc versions, and architecture from the container to inform installation logic, which is then processed by the agent.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection. Ingestion points: JSON outputs from discovery scripts and external Git repository content. Boundary markers: Absent. Capability inventory: docker exec, Bash(*), Write. Sanitization: Absent.

install-stack