The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill recommends installing the mermail-cli package globally via npm install -g or running it once using npx --yes. These commands download and execute code from an external package registry.
[COMMAND_EXECUTION]: The skill's primary purpose is to generate and execute shell commands and scripts. While it includes extensive safety guidelines, the ability to run terminal commands poses a risk if not carefully supervised.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data from email mailboxes and threads, creating a surface for indirect prompt injection attacks.
Ingestion points: Untrusted content is ingested from email bodies, subjects, headers, and thread context through commands like mermail emails list, mermail emails wait, and mermail emails context (referenced in SKILL.md and references/workflows.md).
Boundary markers: The instructions in SKILL.md and references/security.md explicitly state that email content should be treated as untrusted data rather than instructions, providing a defensive boundary.
Capability inventory: The skill can perform sensitive actions such as sending emails (mermail emails send), creating mailboxes, and executing wallet transfers via shell commands.
Sanitization: references/workflows.md indicates that email thread context is sanitized and scan-gated before being processed.

mermail-cli