## 🚀 Default Task Template

Use this template when engaging the Principal Prompt Engineer for a new project or audit.

---

### Prompt: New Agent Architecture Request

```
I need you to design a production-ready AI agent system prompt.

**Agent Concept:** [Describe what the agent should do in 1-2 sentences]

**Target Users:** [Who will interact with this agent?]

**Target Model:** [e.g., Claude 3.5 Sonnet, GPT-4o, or "model-agnostic"]

**Deployment Context:** [Chat UI, API, embedded in product, multi-agent orchestration, etc.]

**Success Criteria:** [How will we know the agent is working? Be specific.]

**Constraints:** [Token budget, latency, compliance requirements, languages, etc.]

**Known Failure Modes:** [Any bad behaviors you've already observed?]

**Deliverable Format:** [Modular Soul files, single system prompt, evaluation suite, full package]
```

---

### Prompt: Prompt Audit Request

```
Audit the following system prompt for production readiness.

**Current Prompt:**
[Paste existing prompt here]

**Observed Issues:** [List any failures, drift, or user complaints]

**Target Model:** [Model name and version if known]

**Use Case:** [What is this agent supposed to do?]

**Evaluation Data:** [Any test cases or logs you can share]

Please provide:
1. Severity-rated findings table
2. Modular refactor recommendation (if applicable)
3. Revised prompt sections with diffs
4. Suggested evaluation test cases (min. 5)
```

---

### Prompt: Quick Optimization

```
Optimize this prompt for [goal: token reduction / better adherence / injection resistance / tone fix]:

[Paste prompt]

Constraints: [Max tokens, must-keep sections, model]

Return: optimized version + change log + estimated token delta
```

---

### Expected Agent Response Structure

For architecture requests, respond with:
1. **Requirements Summary** — Restated understanding + clarifying questions (if any)
2. **Architecture Overview** — PEARL layer diagram + file structure
3. **Modular Files** — Full content for each file
4. **Evaluation Suite** — 5+ test cases with expected behaviors
5. **Deployment Notes** — Versioning, monitoring, and iteration guidance