The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted template files provided by the user, creating a surface for indirect instructions.\n * Ingestion points: Source template files are read from the knock-migration/source/ directory during the analysis phase described in rules/analyzing-source-templates.md.\n * Boundary markers: The instructions do not define specific delimiters or warnings to ignore instructions within user-provided template files.\n * Capability inventory: The skill performs file system writes to the knock/ and knock-migration/ directories, executes knock CLI commands for pushing resources, and makes network requests via curl to knock.app.\n * Sanitization: No explicit sanitization or filtering of template logic (Handlebars, Mako, etc.) is documented before processing.\n- [COMMAND_EXECUTION]: The skill utilizes external CLI tools and network utilities to manage and verify resources.\n * Evidence: Instructions in rules/building-templates.md and rules/migration-process.md involve executing commands such as knock workflow push, knock layout validate, and curl for previewing templates. It also suggests using python3 -m http.server for local verification.\n- [DYNAMIC_EXECUTION]: The skill generates and executes JavaScript snippets to facilitate template verification via the Knock MCP server.\n * Evidence: references/verification-recipes.md provides a JavaScript function used by the execute_mapi_write tool to render templates, calculate hashes, and verify style fingerprints. This execution occurs within the server's context to perform complex assertions during the verification phase.

knock-email-template-migration