Meta-prompts for creating, documenting, and testing other prompts. These tools help with prompt engineering workflows, standardized documentation generation, and security testing against adversarial inputs.
- prompt-documenter.md - Prompt Documenter: Generates standardized documentation for AI prompts with language detection and parameter recommendations
- template-xml.md - XML Template: Basic XML-structured prompt template with role, context, instructions, and example sections
- template-xml-max.md - Advanced XML Template: Extended template with tool integration, multi-format responses, and comprehensive examples
- test-instruction-injection.md - Instruction Injection Tests: Collection of adversarial phrases for testing LLM resistance to prompt injection attacks
- test-instruction-injection-rus.md - Russian version of instruction injection tests for multilingual security assessment
Most prompts suggest:
temperature: 0.2 # Lower creativity for consistent, deterministic outputs in meta-tasks
reasoning_effort: "mediu/high" # Structured analysis and multi-step processing required
verbosity: "low/medium" # Focused outputs without unnecessary elaboration
presence_penalty: 0.3 # For security tests to reduce repetitive attack responses