The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from multiple sources which is then transmitted to an external API, creating a vulnerability to indirect prompt injection.
Ingestion points: Reads data from stdin in _read_stdin(), reads HTML file content via --html in _resolve_html(), and reads arbitrary files for attachments via --attachments in _encode_attachment() within scripts/send.py.
Boundary markers: No technical boundary markers or sanitization logic is implemented for the ingested content. The skill relies on instructional warnings in SKILL.md directing the agent to get explicit user approval before execution.
Capability inventory: The skill combines file-reading capabilities (any file path provided to --attachments or --html) with network-writing capabilities (POST requests to the well-known service api.resend.com as defined in scripts/_resend_utils.py).
Sanitization: The skill performs no sanitization, escaping, or validation of the email body content or attachment data before transmission.

resend