The Agent Skills Directory

Indirect Prompt Injection: The skill provides methods for an agent to ingest and process external data, specifically captured emails, which could contain instructions from untrusted sources.
Ingestion points: The GET /emails and GET /emails/<id> endpoints described in SKILL.md allow for reading email content programmatically.
Boundary markers: No specific delimiters or "ignore" instructions are recommended when the agent processes the retrieved email data.
Capability inventory: The skill has access to Bash for running npx and curl commands.
Sanitization: The instructions do not include steps for sanitizing or validating the html or text content of emails before the agent acts upon them.
Command Execution: The skill utilizes shell commands to manage the local emulator and interact with its API.
Evidence: Uses curl for API requests and npx to launch the emulate service.
External Downloads: The skill relies on external software packages to provide the emulation environment.
Evidence: References the installation and use of @emulators/adapter-next, @emulators/resend, and the emulate package via NPM.

resend