The Agent Skills Directory

[COMMAND_EXECUTION]: The scripts/install-skill.sh script uses shell commands (mkdir, ln, rm) to manage skill directories and create symbolic links. The SKILL.md instructions also utilize git, npm, and pip for repository management and dependency installation.
[EXTERNAL_DOWNLOADS]: The skill's core workflow involves downloading content from external sources, such as GitHub repositories and Gists, to a local environment for assessment and installation.
[REMOTE_CODE_EXECUTION]: The instructions direct the agent to run package managers like npm install or pip install on externally sourced code. This can lead to the execution of arbitrary code via post-install scripts or build processes in those third-party repositories. The skill attempts to mitigate this through a mandatory security assessment step.
[PROMPT_INJECTION]: As a tool designed to ingest and process untrusted external skills (markdown and code), it possesses an indirect prompt injection surface.
Ingestion points: External GitHub repositories, Gists, and URLs processed during the identification step in SKILL.md.
Boundary markers: The skill relies on a structured output from the oss-security-check skill to distinguish between the assessment results and the untrusted content.
Capability inventory: The agent is authorized to clone repositories, run package managers, and create filesystem links across multiple CLI application directories.
Sanitization: The skill implements comprehensive checks for supply chain anomalies, ClickFix-style malware patterns (including macOS Gatekeeper bypasses), and mandates human review through an "Install and review" recommendation for suspicious findings.

install-skill