The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill provides instructions in multiple files (e.g., SKILL.md, hiring/SKILL.md) for the user to execute @coldiq/mcp@latest using npx. This represents the execution of remote code. As the package is owned by the skill author ('cold-iq'), it is documented as a legitimate vendor resource.
[COMMAND_EXECUTION]: The setup instructions involve shell command patterns for setting environment variables and invoking the node package runner, specifically COLDIQ_API_KEY=<key> npx -y @coldiq/mcp@latest.
[INDIRECT_PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection as it ingests and processes untrusted data from the web. Ingestion points: External data is processed via mcp__coldiq__extract_post_engagement and manual plays that scrape content from LinkedIn comments, G2 reviews, and Capterra reviews. Boundary markers: The instructions do not define delimiters or include 'ignore embedded instructions' warnings for the data ingested from these sources. Capability inventory: The agent has the capability to execute MCP tool calls for web searching, company enrichment, and contact discovery based on the ingested content. Sanitization: There is no evidence of filtering, escaping, or validation being applied to external content before it is processed by the agent.

signal-sourcer