The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill references the importation of modules from esm.sh, a well-known service providing Deno-compatible ESM modules.
[COMMAND_EXECUTION]: Instructions include the installation and local execution of the netlify-cli for development and deployment workflows.
[INDIRECT_PROMPT_INJECTION]: The skill describes patterns for handling untrusted client data, creating a potential surface for indirect injection if processed data is subsequently used in sensitive contexts.
Ingestion points: Untrusted data enters the handler through the Request object and Context properties such as cookies, geo, and the request body (e.g., req.json() in handler examples).
Boundary markers: Explicit delimiters or warnings for embedded instructions in processed data are not present in the instructional code examples.
Capability inventory: Handler logic includes the ability to perform fetch operations, issue redirects via Response.redirect, and rewrite or transform responses using context.next and new Response.
Sanitization: Code examples demonstrate basic data processing (e.g., text.toUpperCase()) but do not explicitly focus on sanitization or validation of untrusted input for downstream LLM consumers.

netlify-edge-functions