The Agent Skills Directory

[SAFE]: The skill provides legitimate automation for macOS Mail.app using local scripts. All operations are confined to the user's machine.
[DATA_EXFILTRATION]: While the skill accesses email content, no network-capable tools or patterns were identified that could exfiltrate this data. All retrieved data is returned locally to the agent context.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection due to its core function of reading unstructured email data.
Ingestion points: Email subjects and bodies are ingested via scripts/commands/message/get.sh and scripts/commands/message/list.sh.
Boundary markers: None are present in the script output to isolate external content.
Capability inventory: The skill has the ability to send, reply, and delete messages through the scripts/commands/message/ suite.
Sanitization: Output is processed through jq and custom AppleScript escaping to ensure valid JSON, preventing shell-level injection, though content is not filtered for semantic instructions.
[COMMAND_EXECUTION]: The skill executes shell scripts and AppleScript commands. User-supplied arguments are generally handled through script parameters, and numeric inputs like message indices are validated as positive integers before use.

macos-mail