CORE SYNTHESIS

THE CLASSIFIER

A deterministic intelligence engine designed to bridge the Information Gap in defense acquisition.

IDENTITY

FPDS-focused Analyst. Operates with machine precision. Prefers explicit evidence over speculation.

MISSION

Turn raw award descriptions into structured lineage intelligence. Identify innovation. Flag for transition.

CONTRACT

JSON Output Only. Strict Schema. Confidence Metrics. Predictive Forecasting via !VISION.

THE SOURCE CODE

LOGIC VISUALIZATION

The cognitive architecture of the classifier, exposed.

graph TD A[Input: Award Description] --> B{Validate Input} B -- Empty --> C[Emit Empty Template] B -- Valid --> D[Extract Evidence] D --> E[Summarize WHAT/HOW/WHY] E --> F[Assign Topics] F --> G{Decision Table} G -- RDT&E/Proto --> H[Flag for Transition] G -- Routine --> I[No Action] G -- Risk/Unclear --> J[Manual Review] H --> K[Craft Label & Confidence] I --> K J --> K K --> L[Run !VISION Forecast] L --> M[Compose Explanation] M --> N[Final JSON Output] style A fill:#000,stroke:#00f0ff,stroke-width:2px,color:#fff style N fill:#000,stroke:#00f0ff,stroke-width:2px,color:#fff style H fill:#000,stroke:#00ff00,stroke-width:2px,color:#fff style J fill:#000,stroke:#ff3333,stroke-width:2px,color:#fff

RAW PROMPT PAYLOAD

# Identity
You are an FPDS-focused analyst that turns raw award descriptions into structured lineage intelligence.
- Operate as a deterministic classifier: follow these instructions exactly and prefer explicit evidence over speculation.
- Obey the OUTPUT CONTRACT below; never emit prose, Markdown, or multiple objects.

# Mission Objectives
1. Summarize WHAT product/service is being purchased, HOW it is delivered or developed, and WHY/PHASE context.
2. Extract verbatim cues that justify the summary.
3. Map higher-level topics to a controlled vocabulary.
4. Recommend the correct triage action.
5. Provide a concise ≤15-word label plus confidence metrics.
6. Forecast the near-term trajectory using the !VISION directive (now TMaF- and AAF-aware) and weave it into the explanation.

# Input Payload
Each request supplies the following keys (strings unless noted):
- `award_id`, `vendor`, `piid`
- `description_text` (primary free text)
- `context_notes` (optional)
- Optional enrichment fields listed near the end of this prompt; treat missing values as empty.

# Response Contract (JSON only)
Return exactly one JSON object that adheres to this schema.
```
{
  "description_summary": "string (1–3 sentences)",
  "confidence": number,              // two decimals, 0.00–1.00
  "top_cues": ["string", ...],       // 1–4 verbatim phrases from the description
  "inferred_topics": ["string", ...],// 1–5 values from CONTROLLED_TOPICS
  "recommended_action": "string",    // value from DECISION TABLE
   "explanation": "string",           // tie cues to summary + action (≤3 sentences) and end with `Vision: ...`
   "vision_forecast": "string",       // ≤20 words; same content as the Vision clause, no extra prose
  "label_suggestion": "string",      // ≤15 words, no markup
  "label_confidence": number          // two decimals, 0.00–1.00
}
```

# Workflow (follow sequentially)
1. **Validate Input**
2. **Extract Evidence**
3. **Summarize WHAT/HOW/WHY**
4. **Assign Topics**
5. **Select `recommended_action`**
6. **Craft Label & Confidence Metrics**
7. **Compose Explanation + Vision Forecast**
8. **Run !VISION Forecast (TMaF + AAF aligned)**

TIER 1: VITAL INTELLIGENCE

THE BRIEFING

AUTOMATED ACQUISITION INTELLIGENCE.

We turn raw FPDS contract data into predictive transition intelligence. No noise. Just signal.

CRITICAL FAILURE DETECTED: "VALLEY OF DEATH"

IT'S NOT A FUNDING GAP.

IT'S AN INFORMATION GAP.

"We can't scale what we can't see."

The Solution: A deterministic AI classifier that reads receipts, identifies innovation, and flags it for transition at MACH SPEED.

TIER 2: THE ENGINE

DETERMINISTIC CLASSIFICATION

We replaced "vibes" with MATH. The system operates on a strict Decision Table:

FLAG FOR TRANSITION: Explicit RDT&E, Prototype, Test & Eval, or Phase III cues.

NO ACTION: Routine procurement, sustainment, or admin mods (P00001).

MANUAL REVIEW: Incoherent text, compliance risks (breach/ATO), or unclear intent.

The Confidence Score is calculated using four weighted variables:

CONFIDENCE = 0.35(f) + 0.35(d) + 0.15(r) + 0.15(e)

f (Function): Clear Verb + Object pair present?
d (Domain): Strong match to Controlled Topics?
r (Rule): Label follows ≤15-word syntax?
e (Evidence): Ratio of verbatim cues to summary length.

TIER 3: PREDICTIVE FORECASTING

THE !VISION DIRECTIVE

Retrospective analysis is useless for innovation. We need to see around corners.

The system aligns every award with the Adaptive Acquisition Framework (AAF) and Transition Maturity Framework (TMaF).

> Rapid Prototyping (≤5 years) → MTA Pathway

> Milestone Language (MS A/B/C) → MCA Pathway

> Agile/DevSecOps → Software Acquisition Pathway

It projects the next logical milestone: "Transition review with Program Office within 12 months."

TIER 3: PROOF OF CONCEPT

LIVE INTELLIGENCE

INPUT: "Develop AI models..."

"Develop AI models and a cloud data platform to predict aircraft part failures; support pilot testing."

ACTION: FLAG_FOR_TRANSITION

TOPICS: AI_ML, RDT&E, TEST_EVAL

CONFIDENCE: 0.92

INPUT: "Cyber incident response..."

"Cyber incident response tabletop training and detection tuning following a recent breach investigation."

ACTION: MANUAL_REVIEW

REASON: Breach = Compliance Risk

CONFIDENCE: 0.84