The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute a local Python utility script (scripts/write_script.py) to resolve command templates and subsequently run the designcraft-cli binary to perform document modifications.
[EXTERNAL_DOWNLOADS]: The skill configuration (references/craft-command-evidence.json) references an external portable ZIP archive hosted on GitHub (github.com/storytold/designcraft) which contains the primary CLI tool used by the skill. Additionally, the instructions suggest installing a related skill via the npx skills add command.
[INDIRECT_PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by ingesting untrusted data from design documents and interpolating that data into file-system and layout modification commands.
Ingestion points: Document metadata, including asset IDs, text content, and layout boundaries, is extracted from user-provided files using the document.inspect, links.list, and font.list commands as defined in references/craft-steps.json.
Boundary markers: There are no explicit delimiters or specific instructions for the agent to ignore potentially malicious embedded content within the design document fields.
Capability inventory: The skill possesses capabilities to modify the file system and document structure through links.relink (asset replacement), transform.resize (layout changes), file.exportPdf (file creation), and file.package (file collection).
Sanitization: There is no evidence of validation or sanitization of the properties extracted from external documents before they are passed as arguments to the execution commands.

designcraft-print-preflight