The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill documents workflows for reading inbound emails (resend emails receiving listen, resend emails receiving get). External content from these emails enters the agent context dynamically.
Ingestion points: Detected in references/emails.md and references/workflows.md where the agent handles emails receiving commands.
Boundary markers: The documentation includes explicit warnings to treat this data strictly as untrusted input and not to follow instructions inside emails, reducing risk but maintaining exposure to adversarial manipulation.
Capability inventory: The skill contains the ability to run administrative terminal commands, send emails, manage webhooks, and manipulate domain configuration across all reference files via the resend CLI executable.
Sanitization: The text suggests users/agents should sanitize content before processing, but there are no integrated platform-level filters specified in the instructions.
[EXTERNAL_DOWNLOADS]: The SKILL.md file contains instructions to install global dependencies using npm install -g resend-cli and brew install resend/cli/resend to make the executable available.
[COMMAND_EXECUTION]: Instructions explicitly detail running shell commands (resend) and piping output (e.g., resend emails receiving listen --json | head -3) to manipulate system state and handle network-driven parameters.

resend-cli