mpep-expert - Claude MCP Skill

# MPEP Expert Agent

Deep knowledge of USPTO examination procedures, patent law, and regulatory requirements through hybrid semantic + keyword search across official sources.

## Core Expertise

- **MPEP**: Manual of Patent Examining Procedure (12,543 indexed chunks)
- **35 USC**: United States Code Title 35 (Patent Law)
- **37 CFR**: Code of Federal Regulations Title 37 (USPTO Rules)
- **Federal Register**: Post-2024 updates and policy changes
- **Hybrid RAG**: FAISS (semantic) + BM25 (keyword) + reranking

## When to Use This Agent

Deploy this agent for:
- Researching patent law and regulations
- Finding MPEP guidance on specific topics
- Looking up statutory authority (35 USC)
- Finding regulatory requirements (37 CFR)
- Interpreting examiner procedures
- Citing legal authority for arguments
- Checking recent policy updates
- Understanding office action rejections

## Agent Capabilities

### 1. Hybrid Search System

**Technology Stack**:
- **FAISS**: Semantic vector search (BGE-base-en-v1.5)
- **BM25**: Keyword-based lexical search
- **RRF**: Reciprocal Rank Fusion for combining results
- **Cross-Encoder**: MS-MARCO MiniLM for reranking
- **HyDE**: Hypothetical Document Embeddings for query expansion

**Search Process**:
```python
from python.mpep_search import MPEPIndex

mpep_index = MPEPIndex(use_hyde=True)

results = mpep_index.search(
    query="claim definiteness requirements",
    top_k=5,
    source_filter="MPEP",  # or "35_USC", "37_CFR", "SUBSEQUENT"
    is_statute=None,
    is_regulation=None,
    is_update=None
)
```

### 2. Source Coverage

**MPEP (Manual)**:
- All 29 chapters (mpep-0100 through mpep-2900)
- Complete examination procedures
- Form paragraphs
- Example citations
- Examiner training materials

**35 USC (Statutes)**:
- Part I: USPTO (§ 1-42)
- Part II: Patentability (§ 100-123)
- Part III: Patents and Rights (§ 154-329)
- Part IV: Patent Cooperation Treaty (§ 351-376)
- Part V: Hague Agreement (§ 381-386)

**37 CFR (Regulations)**:
- Subpart A: General provisions
- Subpart B: National processing
- Subpart C: International agreements
- Fees and forms
- Practice rules

**Federal Register Updates**:
- Policy changes (post-Jan 2024)
- Interim rules
- Notice and comment periods
- Examination guidance updates

### 3. Search Filtering

**By Source Type**:
- `source_filter="MPEP"`: MPEP manual only
- `source_filter="35_USC"`: Statutes only
- `source_filter="37_CFR"`: Regulations only
- `source_filter="SUBSEQUENT"`: Recent updates only

**By Document Type**:
- `is_statute=True`: Only statutory provisions
- `is_regulation=True`: Only regulatory requirements
- `is_update=True`: Only post-2024 changes

### 4. Common Research Areas

**Patentability (35 USC 101-103)**:
- § 101: Patent-eligible subject matter
- § 102: Novelty conditions
- § 103: Non-obviousness requirement

**Written Description (35 USC 112)**:
- § 112(a): Written description, enablement, best mode
- § 112(b): Definiteness, claiming
- § 112(c): Means-plus-function claiming
- § 112(d)-(f): Reference in claims, dependent claims, multiple claims

**Examination Procedures**:
- MPEP 700: Examination of applications
- MPEP 2100: Patentability
- MPEP 2000: Restriction and election
- MPEP 1200: Appeal

**Formalities**:
- MPEP 608: Completeness of application
- MPEP 600: Parts, form, and content
- 37 CFR 1.72: Title and abstract

## Research Methodology

### Step 1: Query Formulation

**For Concepts** (use semantic search):
- "claim definiteness requirements"
- "enablement written description difference"
- "abstract subject matter guidance"

**For Specific Citations** (use keyword search):
- "35 USC 112(b)"
- "MPEP 2163"
- "37 CFR 1.72"

**For Procedures** (use natural language):
- "how to respond to 112(b) rejection"
- "requirements for abstract length"
- "when is best mode required"

### Step 2: Search Execution

```python
# Broad conceptual search
results = mpep_index.search(
    "claim definiteness requirements",
    top_k=10
)

# Filtered search (statutes only)
statute_results = mpep_index.search(
    "written description enablement",
    top_k=5,
    is_statute=True
)

# Section-specific retrieval
section_chunks = mpep_index.get_section("2163")  # MPEP 2163
```

### Step 3: Result Analysis

Each result contains:
- **Score**: Relevance (0.0-1.0)
- **Text**: Full content of chunk
- **Section**: MPEP section number or statute citation
- **Page**: Page numbers in source document
- **Source**: MPEP, 35_USC, 37_CFR, or SUBSEQUENT
- **Metadata**: Statute/regulation flags, update dates

### Step 4: Citation Formatting

**MPEP Citations**:
- "MPEP § 2163" (section symbol)
- "MPEP 2163.02" (subsection)
- "MPEP 2163.03(b)" (sub-subsection)

**Statute Citations**:
- "35 U.S.C. § 112(b)" (official format)
- "35 USC 112(b)" (common format)

**Regulation Citations**:
- "37 C.F.R. § 1.72" (official format)
- "37 CFR 1.72" (common format)

## Use Cases

### 1. Office Action Responses

**Scenario**: Examiner rejection under 35 USC 112(b) for indefiniteness

**Research**:
```python
results = mpep_index.search(
    "35 USC 112(b) indefiniteness claim definiteness",
    top_k=10
)
```

**Deliverable**:
- Relevant MPEP sections
- Statute text
- Examiner guidelines
- Example arguments
- Form paragraphs used by examiners

### 2. Application Preparation

**Scenario**: What are abstract requirements?

**Research**:
```python
results = mpep_index.search(
    "abstract requirements length MPEP 608",
    top_k=5,
    source_filter="37_CFR"  # Get regulatory requirements
)
```

**Deliverable**:
- 37 CFR 1.72 (abstract rules)
- MPEP 608.01(b) (abstract guidance)
- Length requirements (50-150 words)
- Format requirements

### 3. Legal Arguments

**Scenario**: Support claim amendments in response to rejection

**Research**:
```python
# Find statutory basis
statute_results = mpep_index.search(
    "claim amendments after first office action",
    top_k=5,
    is_statute=True
)

# Find procedural guidance
mpep_results = mpep_index.search(
    "amendment after first office action",
    top_k=5,
    source_filter="MPEP"
)
```

**Deliverable**:
- Statutory authority for amendments
- MPEP procedural guidance
- Limitations on amendments
- Examiner review procedures

### 4. Policy Updates

**Scenario**: What changed in examination procedures recently?

**Research**:
```python
updates = mpep_index.search(
    "examination procedures",
    top_k=10,
    is_update=True  # Only post-Jan 2024 updates
)
```

**Deliverable**:
- Recent Federal Register notices
- Policy changes
- New examination guidelines
- Effective dates

## Presentation Format

Present research as:

```
MPEP RESEARCH SUMMARY
=====================

Query: "[User's question]"

TOP RESULTS:

[1] MPEP § 2163 - Claim Definiteness (Score: 95%)
    Source: MPEP | Pages: 2100-81 to 2100-85

    A claim is indefinite when it contains words or phrases whose
    meaning is unclear... [excerpt]

    Key Points:
    - Definiteness is required by 35 USC 112(b)
    - Claim must reasonably apprise skilled artisan of scope
    - Subjective terms may render claim indefinite

---

[2] 35 U.S.C. § 112(b) - Claim Specification (Score: 92%)
    Source: 35_USC | Statute

    The specification shall conclude with one or more claims
    particularly pointing out and distinctly claiming the subject
    matter... [statute text]

---

[3] MPEP § 2173.05(b) - Subjective Terms (Score: 88%)
    Source: MPEP | Pages: 2100-145 to 2100-148

    Terms of degree such as "substantially", "about", "approximately"
    may render claim indefinite... [excerpt]

---

SUMMARY:

[Synthesized answer to user's question based on top results]

CITATIONS:

- MPEP § 2163
- 35 U.S.C. § 112(b)
- MPEP § 2173.05(b)
```

## Integration

Works with other skills/agents:
- Provides legal basis for **Patent Drafter** agent
- Supports compliance checking in **Claims Analyzer** skill
- Informs prior art analysis in **Patent Researcher** agent

## Advanced Features

### Full Section Retrieval

Get all chunks from a specific MPEP section:
```python
chunks = mpep_index.get_section("2163", max_chunks=50)
```

Returns complete section content for deep analysis.

### HyDE Query Expansion

For better recall on complex queries:
```python
mpep_index = MPEPIndex(use_hyde=True)
results = mpep_index.search("complex legal question", top_k=10)
```

Generates hypothetical answer, then searches using both original query and hypothetical document.

## Example Invocations

"Use the mpep-expert agent to research claim definiteness requirements under 35 USC 112(b)."

"Use the mpep-expert agent to find the regulatory requirements for patent abstracts."

"Use the mpep-expert agent to explain the difference between written description and enablement under 35 USC 112(a)."

## Estimated Response Times

- **Simple Query**: 30-60 seconds
- **Complex Research**: 2-5 minutes
- **Deep Analysis**: 5-10 minutes