The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes file names, paths, and content retrieved from external platforms (e.g., Webflow, HubSpot, Contentful), which are untrusted sources.
Ingestion points: External data enters the agent's context through tools such as list_integration_files (Connector path) or commands like crowdin app api integration-files (CLI path).
Boundary markers: The skill does not explicitly define unique delimiters or instructions to treat the ingested platform data as untrusted content.
Capability inventory: The skill has the capability to perform network writes back to these external platforms through tools like update_integration_files and update_crowdin_files.
Sanitization: The skill includes clear instructions to use jq --arg when handling paths and names to prevent shell injection, and it mandates that entry IDs be copied "byte for byte" from the platform's own output to maintain integrity.
[EXTERNAL_DOWNLOADS]: The CLI path requires the installation of the Crowdin CLI tool from a well-known package registry.
Evidence: npm install -g @crowdin/cli is listed as a requirement in references/cli.md.
[COMMAND_EXECUTION]: The skill uses a vendor-provided CLI tool to perform operations, involving multiple shell command invocations.
Evidence: references/cli.md details the use of crowdin app list, crowdin app methods, and crowdin app api to interact with the Crowdin platform.

integration-sync