The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill provides instructions for users to install a Model Context Protocol (MCP) server using the command npx -y @coldiq/mcp@latest. This package is a resource managed by the vendor 'Cold-IQ'.
[EXTERNAL_DOWNLOADS]: The skill references external resources and API endpoints hosted on coldiq.com and api.coldiq.com, which are vendor-controlled domains. It also links to educational content on YouTube and Loom, and professional profiles on LinkedIn.
[PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection because it is designed to ingest and process untrusted data from external sources to generate personalized email copy.
Ingestion points: Untrusted data enters the agent context through tools such as mcp__coldiq__fetch_page_content, mcp__coldiq__extract_post_engagement, and mcp__coldiq__find_signals (defined in SKILL.md).
Boundary markers: The instructions lack explicit boundary markers or warnings to the agent to disregard instructions potentially embedded within the external data (SKILL.md).
Capability inventory: The skill's capabilities are focused on text generation and strategy routing; it does not appear to possess tools for high-risk operations like persistent file writing or arbitrary system command execution.
Sanitization: There is no evidence of sanitization or validation logic to filter out malicious payloads or instructions from the retrieved web content or social signals.

cold-email