Prompt Engineering Skills

A collection of model-agnostic prompt engineering skills for OpenCode and Claude Code. Designed for AI engineers who write and iterate on prompts in code — API calls, agent frameworks, system prompts, and multi-step pipelines.

Skills

The collection is structured as one router skill and six focused sub-skills. The router handles every prompt task and loads sub-skills on demand for deeper reference.

Skill	Description
prompt-engineering	Router — entry point for all prompt tasks. Covers the core workflow, model selection guidance, a technique quick-reference table, and a full design checklist and debugging table.
prompt-clarity-and-structure	Vague prompts and inconsistent output. Covers clarity, specificity, structured delimiters, role/persona assignment, few-shot prompting, and output consistency.
prompt-reasoning-and-chaining	Multi-step reasoning and complex workflow decomposition. Covers chain-of-thought levels, structured reasoning, reasoning model caveats, prompt chaining patterns, inter-step validation, and shared concept alignment.
prompt-grounding-and-rag	Hallucinations and RAG pipelines. Covers quote-then-answer, source restriction, confidence calibration, long context handling, lost-in-the-middle, chunking strategies, and reranking.
prompt-agents-and-tools	Tool-using agents and multi-agent systems. Covers tool descriptions, schema design, tool selection, multi-tool orchestration, error handling, context window management, termination conditions, state management, and handoff patterns.
prompt-security-and-production	Production deployment. Covers prompt injection defense, sandwich defense, canary tokens, data-borne injection, provider portability, negative constraint pitfalls, prompt caching, and multimodal prompting.
prompt-evaluation	Measuring and iterating on prompt quality. Covers eval set design, metric selection, LLM-as-judge, A/B comparison, regression testing after model upgrades, and production monitoring.

Installation

Install all skills

OpenCode:

for skill in prompt-engineering prompt-clarity-and-structure prompt-reasoning-and-chaining prompt-grounding-and-rag prompt-agents-and-tools prompt-security-and-production prompt-evaluation; do
  mkdir -p ~/.config/opencode/skills/$skill
  curl -fsSL https://raw.githubusercontent.com/kzhekov/prompt-engineering-skill/refs/heads/main/$skill/SKILL.md \
    -o ~/.config/opencode/skills/$skill/SKILL.md
done

Claude Code:

for skill in prompt-engineering prompt-clarity-and-structure prompt-reasoning-and-chaining prompt-grounding-and-rag prompt-agents-and-tools prompt-security-and-production prompt-evaluation; do
  mkdir -p ~/.claude/skills/$skill
  curl -fsSL https://raw.githubusercontent.com/kzhekov/prompt-engineering-skill/refs/heads/main/$skill/SKILL.md \
    -o ~/.claude/skills/$skill/SKILL.md
done

Install a single skill

Replace <skill-name> with the skill you want (e.g. prompt-grounding-and-rag).

OpenCode:

mkdir -p ~/.config/opencode/skills/<skill-name>
curl -fsSL https://raw.githubusercontent.com/kzhekov/prompt-engineering-skill/refs/heads/main/<skill-name>/SKILL.md \
  -o ~/.config/opencode/skills/<skill-name>/SKILL.md

Claude Code:

mkdir -p ~/.claude/skills/<skill-name>
curl -fsSL https://raw.githubusercontent.com/kzhekov/prompt-engineering-skill/refs/heads/main/<skill-name>/SKILL.md \
  -o ~/.claude/skills/<skill-name>/SKILL.md

For Claude Code, installing into ~/.claude/skills/ makes the skill available globally across all projects. To scope it to a single project, use .claude/skills/<skill-name>/ inside the project root instead.

Skills are picked up automatically on the next session start.

Usage

The prompt-engineering router skill auto-triggers when you ask about:

• Writing or improving prompts
• System prompts or agent instructions
• Few-shot examples, chain-of-thought, prompt templates
• Making an LLM behave differently
• Prompt debugging or evaluation

You can also invoke any skill manually with /prompt-engineering, /prompt-grounding-and-rag, etc.

License

MIT