The Agent Skills Directory

[COMMAND_EXECUTION]: The skill operates by executing the gws CLI binary. It provides a structured way to pass parameters like recipients, subjects, and attachments to the system command.
[INDIRECT_PROMPT_INJECTION]: The skill presents an interface for processing external data via email content flags, which is a potential surface for indirect prompt injection.
Ingestion points: The --subject and --body flags in SKILL.md allow the interpolation of untrusted data into the gws command.
Boundary markers: The instructions do not define specific delimiters or "ignore instructions" markers for the email content.
Capability inventory: The skill has the capability to execute shell commands via the gws binary.
Sanitization: No explicit sanitization or input validation is described for the content passed to the CLI.
[SAFE]: The skill documentation includes a mandatory [!CAUTION] block that identifies the action as a write command and instructs the agent to obtain user confirmation before execution. This mitigation effectively manages the risk associated with the tool's capabilities.

gws-gmail-send