The Agent Skills Directory

[SAFE]: The skill instructions were audited for malicious patterns, and no security issues were found. Its functionality is limited to reasoning and providing architectural advice based on pre-existing local reference files.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes user input from the $ARGUMENTS parameter to determine architectural constraints. While it acts as a data ingestion point, the skill only has permission to read local reference files and invoke search tools. The absence of file-writing or network-access capabilities ensures that any potential injection cannot result in data exfiltration or persistent system changes.\n- [COMMAND_EXECUTION]: The skill utilizes specific qmd CLI tools and MCP search functions to query research data. These tools are restricted to read-only operations for information retrieval and do not provide a path for arbitrary command execution or shell injection.

recommend