Llm Prompt Generation

# LLM Instruction Prompt Generation

**Purpose:** Templates and authoring rules for the paste-ready LLM instruction prompts Plea pairs with every demand and every report.

Plea emits two granularities. Both are mandatory output — not optional.

| Granularity | Where | Purpose |
|-------------|-------|---------|
| Per-request prompt | Embedded inside each `## Request` block (`### LLM Instruction Prompt`) | Hand off a single demand for analysis, design, spec drafting, or prototyping |
| Per-report prompt | At the end of the report (`## LLM Orchestration Prompt`) | Hand off the full demand batch to a single downstream agent |

---

## Action Verbs

One verb per prompt, declared at the top of `# Your task`.

| Verb | When to use | Suggested next agent |
|------|-------------|----------------------|
| `ANALYZE` | Understand scope, root cause, market fit | Field, Compete |
| `PROPOSE` | Generate feature options with hypothesis and KPIs | Spark |
| `DESIGN` | Translate demand into UX flow or interaction model | Vision, Palette, Echo |
| `DRAFT-SPEC` | PRD, user story, or staged spec package | Scribe, Accord |
| `PROTOTYPE` | Working code, mock UI, or runnable demo | Forge, Builder |
| `REFINE` | Iterate on existing demand — add detail, narrow scope, resolve contradictions | Plea (self), Field |

### Default verb by receiving agent

| Receiving agent | Default verb | Why |
|-----------------|--------------|-----|
| Spark | `PROPOSE` | Spark structures feature proposals |
| Scribe | `DRAFT-SPEC` | Scribe writes PRDs and user stories |
| Accord | `DRAFT-SPEC` | Accord builds L0-L3 spec packages |
| Builder / Forge | `PROTOTYPE` | Builder ships code; Forge ships rapid prototypes |
| Rank | `ANALYZE` | Rank quantifies priority — needs analysis input |
| Field | `REFINE` or `ANALYZE` | Field validates synthetic hypotheses |
| Voice | `REFINE` | Voice cross-checks synthetic demand against real feedback |
| Compete | `ANALYZE` | Compete benchmarks against rivals |
| Vision / Palette | `DESIGN` | Design agents need flow framing |

---

## Authoring Rules

- Always quote the user voice verbatim inside the prompt — never paraphrase out the emotion.
- Tag synthetic origin explicitly (`synthetic: true`) so the receiving agent calibrates confidence.
- State the action verb at the top of `# Your task`. One verb per prompt.
- Include acceptance criteria so the receiving agent has a definition of done.
- End with constraints on hypothesis-handling, traceability, and contradiction-preservation.
- In `multi` Recipe output: embed each demand's `engine_concurrence` and calibration tags so downstream agents know whether they are acting on a 3/3-validated demand or a 1/3-divergent hypothesis.

---

## Per-Request Prompt Template

Embedded inside each `## Request` block as `### LLM Instruction Prompt`: