The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes project data and external tool outputs (e.g., from read_project and preview_timeline) to verify edits, creating an ingestion surface for external content.
Ingestion points: Data returned by project inspection tools and metadata from signed frame URLs in SKILL.md.
Boundary markers: None explicitly defined in the instructions to separate data from instructions.
Capability inventory: The skill utilizes file reading (read_project, inspect_item), asset browsing, and file downloading to a temporary directory.
Sanitization: The instructions specifically require shell-quoting URLs before downloading, which is a security best practice to prevent command injection.
[EXTERNAL_DOWNLOADS]: The skill instructs the agent to download signed frame URLs for visual verification when direct inspection is not possible. These URLs are generated by the vendor's MCP server and are used for the primary purpose of the skill.
[COMMAND_EXECUTION]: The skill implies the execution of shell commands (e.g., curl or wget) to download files, specifically emphasizing the use of shell-quoting to ensure safety during the process.

verification