The Agent Skills Directory

[DATA_EXFILTRATION]: The skill facilitates the uploading of local files to a remote server through the mailcoach create-subscriber-import command. By using the @ prefix in the --field parameter, the agent can be instructed to read any file on the local filesystem and transmit it to the configured Mailcoach instance. This legitimate tool feature represents a path for potential data exfiltration if the agent is manipulated into selecting sensitive files like SSH keys or configuration files.
Evidence: Found in SKILL.md and references/command-patterns.md where the syntax mailcoach create-subscriber-import --field csv=@/path/to/file.csv is documented.
[INDIRECT_PROMPT_INJECTION]: The skill operates on data ingested from external sources, creating a vulnerability surface for indirect prompt injection.
Ingestion points: The skill reads data from CSV files during subscriber imports (subscribers.md), retrieves and displays campaign content (campaigns.md), and processes transactional email content (transactional.md).
Boundary markers: No delimiters or specific 'ignore instructions' warnings are provided to help the agent distinguish between data and control instructions within these inputs.
Capability inventory: The skill has powerful capabilities including sending emails to large lists (send-campaign), triggering webhooks (automations.md), and sending transactional messages (send-transactional-mail).
Sanitization: The documentation does not describe any sanitization or validation processes for the data before it is displayed to or processed by the agent.

mailcoach