The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill references a download for a portable binary executable from an external GitHub repository not identified as a trusted vendor. Evidence: references/craft-command-evidence.json specifies a download URL for filmcraft-0.8.0-windows-x64-portable.zip from the storytold/filmcraft repository.
[REMOTE_CODE_EXECUTION]: The skill provides instructions for the agent to download and subsequently execute an external binary tool, which is a high-risk pattern. Evidence: The runtime object in references/craft-command-evidence.json provides the source URL and the executable name (filmcraft-cli.exe) for local execution.
[COMMAND_EXECUTION]: The skill provides detailed instructions and examples for the agent to execute a CLI tool with generated parameters. Evidence: references/craft-methods.md and references/craft-steps.json outline the usage of filmcraft-cli.exe to process project operations defined in JSONL format.
[INDIRECT_PROMPT_INJECTION]: The skill ingests and processes external video media content, creating a vulnerability surface where embedded data (such as metadata or transcriptions) could influence the agent's logic. 1. Ingestion points: The file.import command in references/craft-steps.json is used to load external video files. 2. Boundary markers: No explicit delimiters or instructions are provided to ensure the agent ignores instructions embedded within the imported media. 3. Capability inventory: The skill has the ability to execute CLI commands, read/write files, and import external media. 4. Sanitization: No sanitization or validation of data extracted from imported media files is described in the provided instructions.

filmcraft-short-video