The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides documentation for executing system-level commands through the bsdkrun CLI, including booting OCI images (bsdkrun linux), executing commands in guest environments (bsdkrun exec), and defining provisioning scripts (bsdkrun flavor add --provision).
[EXTERNAL_DOWNLOADS]: The documented tool fetches external assets, such as OCI images from container registries, FreeBSD/NetBSD builds via bsdkrun fetch, and guest agent binaries determined by the BSDKRUN_AGENT_VERSION environment variable.
[DATA_EXFILTRATION]: The bsdkrun ssh setup command is described as accessing the ~/.ssh/ directory to locate and install the user's public keys (id_*.pub) into the virtual machine. While this is a common administrative task for SSH provisioning, it involve access to a sensitive user path.
[PROMPT_INJECTION]: The skill facilitates the ingestion of external data that could lead to indirect prompt injection vulnerabilities if the processed data contains malicious instructions.
Ingestion points: User-supplied OCI image references, guest command arguments, and git repository URLs processed by the linux, freebsd, netbsd, and flavor subcommands.
Boundary markers: The skill documentation highlights the use of the -- separator to distinguish between bsdkrun CLI flags and guest commands.
Capability inventory: The tool possesses capabilities for guest command execution, host-to-guest directory mounting (--mount), network management, and persistent volume modifications.
Sanitization: No explicit sanitization or input validation mechanisms are described for the external inputs handled by the CLI in the provided documentation.

bsdkrun-cli