The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill provides a command to install a remote MCP server from the vendor's domain.
Evidence: claude mcp add --transport http manypi https://mcp.manypi.com/mcp -s user in SKILL.md.
[REMOTE_CODE_EXECUTION]: The skill instructs the user to add an external tool using a remote URL, which allows for the execution of logic hosted on the vendor's infrastructure.
Evidence: The use of claude mcp add pointing to https://mcp.manypi.com/mcp.
[INDIRECT_PROMPT_INJECTION]: The skill is inherently designed to ingest and act upon untrusted data from external parties (email replies).
Ingestion points: Prospect replies to cold emails (e.g., "Interested", "Questions", "Objections") as described in the Classify first section of SKILL.md.
Boundary markers: The instructions lack specific delimiters or warnings to treat processed email content as untrusted data.
Capability inventory: The skill references the manypi tool, which is described as having capabilities to route replies, monitor mailboxes, and manage automated email sequences.
Sanitization: There is no evidence of content sanitization or instruction to filter malicious patterns within email replies before processing.

reply-handling