from OpenClaw Master Skills2,002
A practical collection of prompt engineering techniques and patterns for building reliable, controllable LLM prompts: CoT, few-shot, structured outputs, templat
This skill is a comprehensive patterns guide for prompt engineering. It documents concrete techniques — few-shot selection, chain-of-thought, structured JSON outputs, template systems, progressive disclosure, and error recovery — with code examples and best practices for production use. It's designed to make prompts more reliable, efficient, and easy to maintain.
Use this skill when designing or debugging prompts for production LLM systems: building structured output pipelines, reducing hallucinations, implementing self-verification, or constructing reusable prompt templates. Helpful for engineers, ML engineers, and prompt authors.
Agent-agnostic patterns that work with Claude, Anthropic, OpenAI, LangChain-based chains, and other LLM runtimes supporting system prompts and structured output modes.
A comprehensive prompt engineering patterns skill covering CoT, few-shot, structured outputs, templates, and system prompt design. The bundled script optimize-prompt.py failed on import due to missing numpy dependency. SKILL.md is well-written with good examples but is mostly reference content rather than actionable automation. The script uses a ThreadPoolExecutor without proper cleanup in the optimize() path and lacks type hints on some methods.
numpyNo security concerns — no network calls to unknown hosts, no credential handling, no destructive commands. The script is a local prompt optimization tool using a pluggable LLM client. Security deductions: -12 for unquoted/format-string user input in prompt_template.format(**test_case.input) which could cause KeyError/AttributeError crashes but not injection in this context. Code quality is decent but not exceptional: good structure, clear variable names, but missing numpy is a hard blocker and accuracy metric is simplistic word-overlap. Architecture: frontmatter present, SKILL.md is thorough but monolithic — no references/ directory, scripts/ separated but only one script. Usefulness: prompt engineering is a broadly relevant topic but this skill is more of a reference guide than an automation tool — the script is a demo, not production-ready.
