ai-prompt-engineeringlisted

**CONSUMES**: User prompt engineering task + optional existing system prompts or eval dataset **PRODUCES**: Tested prompt suite + optimization results + CI/CD gate configuration # AI Prompt Engineering Capability Pack **Version**: 1.0.0 **Compatibility**: Claude Code (Phase 1); Codex / Cursor / Gemini in Phase 3 **License**: Apache 2.0 — see LICENSE-ATTRIBUTION.md for source credits --- ## What This Pack Does AI agents can write prompts. What they cannot do: - Test prompts against regression suites before deploying - Diagnose *why* a prompt hallucinates or drifts format - Set up CI/CD pipelines that block bad prompts from shipping - Handle model updates that silently break existing prompts This pack encodes the **production prompt lifecycle** — 4 phases with specific CLI tools at each phase — that most prompt engineering guides skip. **This pack teaches "how to run prompts in production"**, not "how to write a prompt." The Anthropic tutorial covers writing. This covers testing, versioning, drift detection, CI/CD. **Pack = prompt production knowledge. Your workflow system = process constraints. No overlap.** --- ## Contents / Navigation Index | Section | Loads | Use when | |---------|-------|----------| | Step 0: Context Detection Router | this file | Always — entry routing | | Phase 1: Write | `references/phase1-write.md` | Designing a new prompt | | Phase 2: Test | `tools/selection-matrix.md`, `tools/promptfoo-starter.yaml` | Building a regression/eval suite | | P