Universal prompt engineering skill for AI coding agents. Works with any agent that supports the skills concept -- Claude Code, Cursor, Gemini CLI, Codex, Goose, Windsurf, Roo Code, Cline, and 40+ other agents.
Provides structured techniques, audit workflows, and reference materials for crafting, reviewing, and optimizing prompts for any LLM.
- Craft prompts -- XML-tagged structure, output control, scope constraints, uncertainty handling, long-context grounding
- Audit/review prompts -- 8-dimension checklist (clarity, structure, safety, hallucination, context, maintainability, model fit, eval readiness) with severity ratings and fix references
- Agentic prompts -- tool usage rules, user update patterns, self-check for high-risk outputs
- Structured extraction -- schema-based data extraction with validation
- Prompt migration -- cross-model migration checklist
| Category | Files |
|---|---|
| Fundamentals | Prompting introduction, techniques catalog (17 techniques), risks & misuses |
| Model-specific | Claude 4.x and Fable 5, GPT-5 through GPT-5.6 Sol, Gemini 2.5/3/3.1 |
| Mistakes & anti-patterns | Hallucinations, structural fragility, context rot, prompt debt, security vulnerabilities |
| Failure analysis | 18-category taxonomy with minimal reproducible prompts, risk scoring, case studies |
| Evaluation | Metrics, CI gating, red-teaming workflows, tooling ecosystem |
| Audit | 43-check prompt audit checklist across 8 dimensions |
npx skills add CodeAlive-AI/ai-driven-development@prompt-engineering -g -yCopy the prompt-engineering directory (containing SKILL.md and references/) into your agent's skills folder:
| Agent | Path |
|---|---|
| Claude Code | ~/.claude/skills/prompt-engineering/ |
| Cursor | ~/.cursor/skills/prompt-engineering/ |
| Gemini CLI | ~/.gemini/skills/prompt-engineering/ |
| Codex | ~/.codex/skills/prompt-engineering/ |
| Windsurf | ~/.codeium/windsurf/skills/prompt-engineering/ |
| Goose | ~/.config/goose/skills/prompt-engineering/ |
| Roo Code | ~/.roo/skills/prompt-engineering/ |
For the full list of 42 supported agents and their paths, see skills.sh.
Once installed, the skill activates automatically when you ask your agent to:
- "improve this prompt"
- "write a system prompt"
- "audit this prompt"
- "review my prompt"
- "optimize my instructions"
- "help me prompt engineer"
Or invoke directly: /prompt-engineering
MIT