The Agent Skills Directory

[PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection because it is designed to read and process untrusted email content from external sources.
Ingestion points: The references/get-mail.md and references/search-mail.md documents describe processes for retrieving email subjects, bodies, and attachment content which are controlled by external senders.
Boundary markers: The skill includes explicit security rules in references/security.md and references/get-mail.md that instruct the agent to treat email content as data rather than instructions and provides a mandatory warning string for the user if suspicious commands are detected.
Capability inventory: The skill allows for sending, replying to, and forwarding emails (wecom-cli mail send), downloading media files (wecomcli-media media download), and writing to local storage using the Write tool.
Sanitization: There are explicit instructions to ignore instructions embedded in email bodies, detect social engineering features (e.g., mismatched domains, urgent requests for sensitive data), and alert the user to check the source before acting on the content.

wecomcli-email