The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes administrative tools such as apt, dnf, snap, and systemctl to manage system software and services. These operations are essential to the skill's purpose and are documented with safeguards such as configuration snapshots and transaction logging.
[EXTERNAL_DOWNLOADS]: The documentation provides instructions to fetch GPG signing keys from well-known services, specifically NodeSource and HashiCorp, to secure third-party repository communications. These downloads target official, recognized domains and are used for cryptographic verification.
[PRIVILEGE_ESCALATION]: The skill requires administrative privileges (via sudo) for installing packages and modifying system configurations. It also involves using chmod to make AppImage files executable. These are standard administrative actions for Linux system management.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data in the form of package names and repository definitions provided by the operator or the host environment.
Ingestion points: Repository URLs and package name inputs processed by the agent (SKILL.md).
Boundary markers: Workflows mandate the use of simulation flags (e.g., --dry-run, -s) to allow the operator to inspect intended changes before execution.
Capability inventory: The skill has access to package managers and system service controllers (systemd).
Sanitization: The skill relies on human-in-the-loop verification of transaction simulations to identify unintended consequences.

linux-package-management