The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The code examples for handling file operations in src/routes/api/files/[name]/+server.ts exhibit a path traversal vulnerability.
The GET and PUT handlers interpolate the params.name variable directly into file system paths (e.g., ./data/${params.name}) using Bun.file() and Bun.write().
This allows an attacker to supply a malicious filename containing directory traversal sequences (such as ../../) to read or overwrite files outside the intended directory.
Ingestion points: The URL parameter params.name and the request body request.text() in src/routes/api/files/[name]/+server.ts serve as entry points for untrusted data.
Boundary markers: There are no markers or instructions provided to the agent to treat this data as untrusted or to enforce boundaries.
Capability inventory: The skill demonstrates both file reading (Bun.file) and file writing (Bun.write) capabilities within the same module.
Sanitization: The provided code snippets do not include any logic to sanitize, escape, or validate the path before execution.

bun-sveltekit