The Agent Skills Directory

[COMMAND_EXECUTION]: The script executes local system commands to retrieve credentials.\n
Evidence: scripts/email_client.py uses subprocess.run(['pass', 'show', path]) to fetch passwords from the local password store.\n- [DATA_EXFILTRATION]: Sensitive user data is stored locally and potentially sent to external targets.\n
Evidence: scripts/check_email.py and scripts/email_client.py store fetched email subjects and bodies in cache/emails.json without encryption.\n
Evidence: config/accounts.json contains a hardcoded Telegram chat ID (notification_target) used for notifications, which may send sensitive email summaries to an external service.\n- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection via incoming emails.\n
Ingestion points: scripts/email_client.py (via fetch_unread and fetch_recent) reads untrusted email content from external senders.\n
Boundary markers: Absent. No delimiters or instructions are used to separate email content from system instructions in scripts/reply_draft.py.\n
Capability inventory: The skill has the ability to send emails (scripts/email_client.py) and read from a local password store.\n
Sanitization: Absent. The skill does not sanitize or filter email content before processing it for summaries or reply generation.

email-manager