The Agent Skills Directory

[DATA_EXFILTRATION]: Exposure of personal identifiable information and sensitive email data within the skill's distributed files.\n
The file config/accounts.json contains specific, hardcoded email addresses (2067089451@qq.com, aaqwqaa68@gmail.com).\n
The file cache/emails.json contains cached email metadata including subjects and sender information from real-world accounts, which constitutes a data leak in a skill distribution context.\n- [COMMAND_EXECUTION]: Local command execution for credential management.\n
The script scripts/email_client.py uses subprocess.run to call the pass utility to retrieve account passwords, which is a local command execution of a system tool.\n- [PROMPT_INJECTION]: Vulnerability to indirect prompt injection from untrusted external data.\n
Ingestion points: Email subjects and bodies are fetched from remote IMAP servers in scripts/email_client.py.\n
Boundary markers: The skill lacks delimiters or instructions to prevent the AI from obeying commands embedded in email content during summarization or drafting.\n
Capability inventory: The skill can send emails via SMTP (scripts/send_email.py) and perform AI-driven content generation.\n
Sanitization: No sanitization or filtering is applied to the email content before it is processed by the AI.

email-manager