The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill facilitates the connection to an MCP server hosted at https://api-mail.adapty.io/mcp. This is a vendor-controlled domain (adaptyteam) required for the skill's primary functionality.
[COMMAND_EXECUTION]: The documentation provides shell commands (e.g., claude mcp add, codex mcp login) to help users configure the MCP server in their local environment. These are standard setup procedures for the supported clients.
[INDIRECT_PROMPT_INJECTION]: The skill processes data from external sources, such as App Store or Google Play URLs via brand_source_add, and user-provided instructions to generate email sequences. This creates an attack surface where malicious metadata in a store listing could attempt to influence the agent.
Ingestion points: brand_source_add (fetches store listing data), campaign_generate (processes user-provided campaign goals).
Boundary markers: The skill does not use specific technical delimiters but enforces a strict "Review" and "Ask before anything" protocol to ensure the user audits generated content.
Capability inventory: container_save (deploys live email flows), campaign_send_test (sends test emails), domain_delete (removes infrastructure), profile_suppress (modifies user data).
Sanitization: Relies on mandatory human verification blocks defined in the instructions before any write operations occur.
[SAFE]: The skill includes explicit instructions never to ask for or read user passwords or tokens. It also mandates a confirmation block for any action that sends emails to real people or cannot be undone, significantly mitigating the risk of accidental or malicious automated actions.

adapty-mail