The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes shell commands to interact with the file system and perform media analysis. This includes standard utilities like mediainfo, file, and ls.
[COMMAND_EXECUTION]: The installation scripts scripts/install.sh and scripts/install.ps1 attempt to install system packages via apt-get, dnf, pacman, and apk. The PowerShell script explicitly employs sudo to escalate privileges during the installation process, which represents a security risk if executed without oversight.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection (Category 8) due to its ingestion of untrusted metadata from external media files. 1. Ingestion points: The mediainfo command reads metadata from files provided in the agent's environment (SKILL.md, references/advanced-usage.md). 2. Boundary markers: The skill lacks any delimiters or instructions to the agent to treat the metadata as untrusted data. 3. Capability inventory: The skill has access to command execution and file system manipulation. 4. Sanitization: There is no evidence of filtering or validation of the metadata output before it is returned to the agent context. If a file contains malicious instructions in fields such as Title or Comment, the agent may inadvertently follow them.

mediainfo