The Agent Skills Directory

[COMMAND_EXECUTION]: The skill relies extensively on executing the inboxd CLI tool to perform all operations, including reading, searching, analyzing, deleting, and sending emails.
[EXTERNAL_DOWNLOADS]: The skill requires the installation of an external Node.js package from the npm registry using npm install -g inboxd. This package is authored by the same vendor as the skill.
[PROMPT_INJECTION]: The skill is susceptible to Indirect Prompt Injection due to its core function of processing untrusted data from an external source (Gmail).
Ingestion points: Untrusted content enters the agent context through inboxd analyze and inboxd read commands, which fetch email bodies, snippets, and subject lines.
Boundary markers: The instructions do not define specific delimiters or "ignore embedded instructions" warnings to separate email content from the agent's internal logic during analysis.
Capability inventory: The agent has high-impact capabilities including deleting emails (inboxd delete), sending new emails (inboxd send), replying to messages (inboxd reply), and archiving content (inboxd archive).
Sanitization: There is no evidence of content sanitization or filtering to prevent malicious instructions embedded in emails from influencing the agent's behavior.

inbox-assistant