The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/email_client.py uses subprocess.run to call the pass utility for retrieving email credentials. This is a controlled execution of a local system tool and follows the documented security model for the skill.
[DATA_EXFILTRATION]: The skill is configured to send notifications to a specific Telegram chat ID (-1003824568687) as part of its core functionality. While this involves transmitting data to an external service, it is a documented and intended feature for user alerts.
[PROMPT_INJECTION]: The skill processes untrusted email subjects and bodies to generate AI summaries and reply drafts, creating a surface for indirect prompt injection.
Ingestion points: Email data is ingested from external servers by scripts/check_email.py and scripts/reply_draft.py.
Boundary markers: No explicit delimiters or instructions are used to isolate untrusted email content from the AI instructions.
Capability inventory: The skill possesses network access (IMAP/SMTP) and local file system access for caching and configuration.
Sanitization: Incoming email content is truncated to 1000 characters to mitigate the impact of malicious payloads.

email-manager