The Agent Skills Directory

[PROMPT_INJECTION]: The skill maintains an attack surface for indirect prompt injection because it ingests and processes untrusted user-supplied files (manifests and JavaScript code) as part of its core development and validation workflow.
Ingestion points: Reads manifest.json, server.js, and app.js from the local workspace to perform validation, refactoring, and auto-fixes.
Boundary markers: Includes specific security rules in rules/security.mdc that enforce input validation and the use of allowlists for SMI operations.
Capability inventory: The skill has access to powerful tools including shell (running fdk validate, node, npm), read, and write for app creation and testing.
Sanitization: Instructions in rules/security.mdc mandate string sanitization and the rejection of malformed or unexpected data before processing.
[EXTERNAL_DOWNLOADS]: The skill's templates and documentation reference various external assets and scripts necessary for app development.
Fetches the Freshworks Crayons component library and SDK from official CDNs (e.g., cdn.jsdelivr.net, static.freshdev.io, unpkg.com).
References official Microsoft Graph, Slack, and Stripe API endpoints in integration playbooks.
These downloads originate from well-known services or the official vendor domain and are documented as standard platform requirements.

fw-app-dev