The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill references external documentation from the KDE organization's Kdenlive repository on GitHub for workflow guidance. These are informational references to a well-known open-source project and do not involve executing untrusted remote code.
[COMMAND_EXECUTION]: The skill instructs the agent to use a local tool, filmcraft-cli, for video processing tasks. The instructions emphasize safe command construction by using argument lists (argv) to prevent shell-related vulnerabilities and require probing input files for verification before execution.
[INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface where it processes metadata from user-supplied media files to derive parameters for video editing commands.
Ingestion points: Metadata from files like input/action.avi is read into the agent's context using the filmcraft-cli probe command (documented in references/craft-methods.md).
Boundary markers: Not explicitly defined; however, the skill mandates the use of structured JSON for command parameters, which helps delimit data from instructions.
Capability inventory: The skill enables execution of filmcraft-cli with various parameters and file system writes to the output/ directory.
Sanitization: The instructions specifically direct the agent to use argument vectors for CLI calls to mitigate command injection risks associated with untrusted filenames or metadata.

filmcraft-freeze-explainer