The Agent Skills Directory

DATA_EXFILTRATION (MEDIUM): Scripts including poffice_docs.py, poffice_mail.py, and poffice_paperless.py perform network requests to administrative endpoints with verify=False. This bypasses SSL/TLS certificate validation, allowing attackers on the network to intercept sensitive administrative API keys and session tokens.
COMMAND_EXECUTION (MEDIUM): The scripts/poffice_master_admin.py file executes local Python scripts using subprocess.check_output. It constructs argument lists using values (email, name, password) retrieved from an external Neon PostgreSQL database. If the database content is compromised, it could lead to command argument injection within the orchestration flow.
EXTERNAL_DOWNLOADS (LOW): The scripts/poffice_docs.py file provides functionality to download files from a remote Seafile instance to the local environment.
PROMPT_INJECTION (LOW): The skill is vulnerable to Indirect Prompt Injection (Category 8).
Ingestion points: scripts/poffice_imap.py (reads email subjects and bodies) and scripts/poffice_docs.py (downloads cloud files).
Boundary markers: None. There are no instructions to the agent to ignore or sanitize instructions embedded in retrieved data.
Capability inventory: The skill has broad permissions to create mailboxes, manage cloud users, and send emails.
Sanitization: No validation or sanitization is performed on data retrieved from external email or document sources before it enters the agent's context.

poffice-admin