The Agent Skills Directory

[COMMAND_EXECUTION]: The skill includes a Python helper script that executes the HandBrakeCLI system utility to perform media processing.
Evidence: scripts/handbrake.py uses subprocess.run() across multiple subcommands like encode, batch, and inspect.
Security: Commands are constructed as lists rather than shell strings, which effectively prevents shell injection vulnerabilities.
[EXTERNAL_DOWNLOADS]: The documentation provides instructions for users to download and install the HandBrake CLI.
Evidence: SKILL.md references installation via brew, apt, the official HandBrake website, and a well-known Docker image (jrottenberg/handbrake).
Context: These references target official repositories and established technology services.
[PROMPT_INJECTION]: The skill's inspection feature processes external file metadata, which is a potential vector for indirect prompt injection.
Ingestion points: scripts/handbrake.py reads output from HandBrakeCLI --scan and returns it to the agent context.
Boundary markers: The skill does not currently use specific delimiters or instructions to isolate this untrusted metadata.
Capability inventory: The skill has the ability to execute commands and write files to the local system.
Sanitization: While the script prevents command injection during execution through its use of list-based arguments, it does not sanitize the resulting metadata strings before they are presented to the agent.

media-handbrake