The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill defines a process for triaging and classifying untrusted external data in the form of email replies. This creates an attack surface where malicious instructions could be embedded in the replies to influence the agent's behavior during classification or routing.
Ingestion points: Incoming email replies processed for triage (referenced in SKILL.md).
Boundary markers: None specified in the triage workflow description.
Capability inventory: The skill utilizes an MCP server for outreach and reply management, which involves reading external data and potentially performing automated responses.
Sanitization: The instructions suggest periodic spot-checks of automated classification accuracy but do not detail specific sanitization or filtering of reply content.
[REMOTE_CODE_EXECUTION]: The skill provides a command to install an MCP (Model Context Protocol) server from https://mcp.manypi.com/mcp. This server extends the agent's environment to support the ManyPI outreach and triage workflow.
[EXTERNAL_DOWNLOADS]: The installation command connects to mcp.manypi.com to fetch the MCP server configuration. This is a vendor-owned resource used for the skill's core functionality.

high-volume-reply-triage