The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it instructs the agent to scrape content from untrusted external websites and use it in subsequent reasoning steps.
Ingestion points: resources/templates/beginner-workflow.md explicitly directs the agent to visit a /company_domain and summarize its content.
Boundary markers: No delimiters or protective instructions are used to separate the scraped untrusted content from the agent's instructions.
Capability inventory: The skill has the ability to read files, search the file system, and use MCP tools (mcp__coldiq__*) to search and enrich data.
Sanitization: There is no evidence of filtering or sanitizing the scraped website content before it is processed by the LLM.
[EXTERNAL_DOWNLOADS]: The skill provides instructions to download and execute external code.
Evidence: npx -y @coldiq/mcp@latest is recommended for installation in SKILL.md.
Note: The package is owned by the vendor 'Cold-IQ'.
[EXTERNAL_DOWNLOADS]: The skill includes a reference to a domain flagged as malicious by automated reputation scanners.
Evidence: saasmag.com is listed in resources/list-building-advanced/list-building-directories.md as a source for SaaS company lists.

list-building