The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external sources, specifically email threads and calendar events, via the Superhuman Mail MCP server. This creates a surface for indirect prompt injection, where malicious instructions hidden in an email could attempt to influence the agent's behavior during the summarization or triage process.
Ingestion points: SKILL.md (calls to Superhuman_Mail.list_threads and Superhuman_Mail.query_email_and_calendar).
Boundary markers: The skill does not define explicit delimiters or instructions to ignore embedded commands within the processed email content.
Capability inventory: SKILL.md includes capabilities to draft replies (Superhuman_Mail.create_or_update_draft) and archive threads (Superhuman_Mail.update_thread).
Sanitization: No explicit sanitization or filtering of email body content is defined in the instructions.

morning-briefing