The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes untrusted external data from email bodies and meeting transcripts, creating a surface for indirect prompt injection.
Ingestion points: External data enters the agent context via email content retrieved through the gog CLI and meeting transcripts fetched via mcporter.
Boundary markers: The instructions do not specify the use of delimiters or clear separation between system instructions and the untrusted external content.
Capability inventory: The skill has permission to execute shell commands (Bash), access the filesystem (Read, Write, Glob, Grep), and create email drafts.
Sanitization: There is no mention of sanitizing or validating the content of emails or transcripts before they are processed by the agent.
[COMMAND_EXECUTION]: The skill performs shell command execution to interface with external tools and local scripts.
It executes python3 to run logging and auditing scripts (skill_log.py, audit_log.py).
It utilizes CLI tools like gog for Gmail interactions and mcporter for Grain transcript access.
[DATA_EXPOSURE]: The skill accesses a local configuration file (~/executive-assistant-skills/config/user.json) containing sensitive user information.
This file includes the user's primary/work emails, scheduling preferences, and signatures, which are then used across the skill's execution.

email-drafting