senior-prompt-engineer

What This Skill Does

The Senior Prompt Engineer skill is a sophisticated toolkit designed for developers and AI practitioners to refine, optimize, and evaluate LLM interactions. It acts as an expert consultant for prompt engineering patterns, structured output design, and agentic system architecture. By utilizing this skill, users can systematically reduce token costs, improve response consistency, and measure the effectiveness of their RAG pipelines. Whether you are building complex chain-of-thought workflows or simple few-shot templates, this skill provides the CLI-based scripts necessary to analyze, benchmark, and iterate on AI configurations effectively.

Installation

To integrate this skill into your environment, run the following command in your terminal: clawhub install openclaw/skills/skills/alirezarezvani/senior-prompt-engineer Ensure your local environment is configured to access the OpenClaw plugin registry. For detailed integration, refer to the source repository at openclaw/skills.

Use Cases

• Prompt Optimization: Automatically detect ambiguity, redundancy, and structure gaps in system prompts to reduce latency and costs.
• Agent Orchestration: Visualize complex multi-agent workflows to ensure logic flow is optimal and scalable.
• RAG Validation: Rigorously test your retrieval-augmented generation systems by measuring context relevance against ground-truth question sets.
• Structured Output Enforcement: Generate and test few-shot example sets to ensure consistent JSON or XML outputs from LLMs.

Example Prompts

1. "Analyze my system_prompt.txt file and suggest 3 ways to reduce token usage while increasing the clarity of the output constraints."
2. "I need a few-shot example template for a customer support agent. Generate 5 variations that handle edge-case complaints."
3. "Evaluate the performance of my RAG pipeline using the data_contexts.json and eval_questions.json files to check for faithfulness scores."

Tips & Limitations

• Performance: The prompt optimizer works best with standard English inputs; verify outputs for non-standard languages.
• Security: Be cautious when passing sensitive internal documents as context files to the RAG evaluator.
• Cost: While this skill optimizes for token usage, ensure you have sufficient API credits for the models being used (e.g., GPT-4) during evaluation runs.
• Context: Always provide a representative sample of 'ground truth' data for the RAG evaluator to ensure accuracy in metrics.