meta-prompting-suite

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (medium risk: 0.30). In patterns/alchemical-agent.md, /alchemical-agent first generates a subagent definition from the user-supplied <role-name> and $ARGUMENTS via an internal meta-design pass, then writes the resulting agent prompt to ~/.claude/agents/ and hands it to /agents, so the LLM directly ingests outsider-authored free text (the arguments) at runtime.