The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external data such as email content and user instructions, creating a surface where malicious instructions could potentially influence the agent. 1. Ingestion points: The skill ingests user input for recipients, subjects, and email bodies as described in SKILL.md. 2. Boundary markers: The instructions do not specify explicit delimiters for separating user-provided content from agent instructions. 3. Capability inventory: The skill utilizes the gmail CLI and standard shell tools like cat for file operations. 4. Sanitization: It recommends using temporary files and quoted heredocs (<<'EOF') to mitigate shell injection risks and emphasizes a draft-based workflow allowing for human verification before any email is sent.
[COMMAND_EXECUTION]: The skill instructions utilize the gmail command-line utility and basic shell commands like cat. These operations are transparently documented and aligned with the primary purpose of managing email drafts.

gmail-draft