The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill provides multiple templates that process external data without implementing sufficient sanitization, creating an attack surface for indirect prompt injection and XSS.
Ingestion points: API responses in BlogPostPage and form data in ValidatedPostForm.
Boundary markers: Absent; data is directly interpolated into components.
Capability inventory: Templates demonstrate network fetching, file writing, and server-side redirects.
Sanitization: Absent; templates/app-router-async-params.tsx and templates/server-actions-form.tsx utilize dangerouslySetInnerHTML to render content directly from external API responses.
[DYNAMIC_EXECUTION]: The file upload template in templates/server-actions-form.tsx contains a path traversal vulnerability that could be exploited to overwrite files on the server.
Evidence: The uploadImage server action constructs a file path using a user-controlled filename: const filename = \${Date.now()}-${file.name}`; const path = join(process.cwd(), 'public', 'uploads', filename); await writeFile(path, buffer);`.
Vulnerability: While the timestamp prefix provides a minor hurdle, it does not prevent path traversal if file.name contains .. sequences, potentially allowing an attacker to write files outside of the intended uploads directory.
[COMMAND_EXECUTION]: The skill includes a utility script and documentation instructing the execution of shell commands and package managers.
Evidence: scripts/check-versions.sh uses node -p to parse local package files. The README.md and commands/init.md instruct users to run npx commands for setup and migration, specifically targeting @next/codemod and create-next-app.

nextjs