The Agent Skills Directory

[PROMPT_INJECTION]: Indirect Prompt Injection Surface. The skill fetches untrusted external data which could contain malicious instructions designed to exploit the agent's broad toolset. (1) Ingestion points: Uses mcp__firecrawl__firecrawl_search and WebFetch to gather requirements. (2) Boundary markers: Absent; the skill does not specify delimiters or instructions to treat fetched data as data only. (3) Capability inventory: Authorized to use Read, Write, Edit, and Bash tools. (4) Sanitization: Absent; no evidence of input validation or sanitization for external content.
[COMMAND_EXECUTION]: The skill uses the Bash tool for legitimate development tasks, such as creating directories and initializing project files for new agents.
[EXTERNAL_DOWNLOADS]: The skill references and provides installation examples for official Model Context Protocol (MCP) packages, which are well-known and trusted development resources.

agent-creator